Penguin_species.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "bbc5d4a4-8ff0-4467-8305-f15cc36241de",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "# Project 1: Predicting Penguin species "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9772053c-7c30-49a1-8381-370f7c0dc07d",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### by: Elmira Onagh\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b6256cea-40bc-4dd1-a5f7-977b7ce1db88",
   "metadata": {
    "tags": []
   },
   "source": [
    "### Citation 1"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d4c736cb-494a-4d66-8c26-feeac6d1c791",
   "metadata": {},
   "source": [
    "Gorman KB, Williams TD, Fraser WR (2014) Ecological Sexual Dimorphism\n",
    "and Environmental Variability within a Community of Antarctic Penguins (Genus Pygoscelis). PLoS\n",
    "ONE 9(3): e90081. doi:10.1371/journal.pone.0090081\n",
    "\n",
    "the data is taken from [here](https://github.com/allisonhorst/palmerpenguins/blob/master/README.md)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e840b269-c2a5-4163-8faf-1ab4ef8b1ac5",
   "metadata": {
    "tags": []
   },
   "source": [
    "### Citation 2"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "52d251fc-bc56-46cb-9847-9b8a745cc0de",
   "metadata": {},
   "source": [
    "The codes used for formatting purposes are taken from the material covered in the lecture by Prof. Archit Garg"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9cdb555b-90de-4956-ae5c-b26dac788b7d",
   "metadata": {},
   "source": [
    "### Citation 3"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f1195452-41a7-4c87-bea8-9b7cfd50a56c",
   "metadata": {},
   "source": [
    "Learning curve sections: citation: https://www.dataquest.io/blog/learning-curves-machine-learning/"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "25e6491e-b942-43d7-ba03-dcb1f0776741",
   "metadata": {
    "tags": []
   },
   "source": [
    "### Variables"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "02bca6ec-2d7d-43f9-ad14-74688d0d8f03",
   "metadata": {},
   "source": [
    "- **species** (str) - penguin species (Chinstrap, Adélie, or Gentoo) --> Target variable\n",
    "- **culmen_length_mm** (float) - culmen length in mm  --> continuous \n",
    "- **culmen_depth_mm** (float) - culmen depth in mm  --> continuous \n",
    "- **flipper_length_mm** (float) - flipper length in mm --> continuous \n",
    "- **body_mass_g** (float) - body mass in g  --> continuous\n",
    "- **island** (str) - island name (Dream, Torgersen, or Biscoe) in the Palmer Archipelago (Antarctica) -->categorical\n",
    "- **sex** (str) - penguin sex  --> categorical"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "83755fea-5187-4f26-a3b4-7cf8c0546ea5",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "## **Part I: Exploratory Data Analysis**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bf000f58-330d-41c6-8961-f786a1887473",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### Reading data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "7738e10d-7c60-452b-a53c-47b286aea2b0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>island</th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "      <th>sex</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>39.1</td>\n",
       "      <td>18.7</td>\n",
       "      <td>181.0</td>\n",
       "      <td>3750.0</td>\n",
       "      <td>MALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>39.5</td>\n",
       "      <td>17.4</td>\n",
       "      <td>186.0</td>\n",
       "      <td>3800.0</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>40.3</td>\n",
       "      <td>18.0</td>\n",
       "      <td>195.0</td>\n",
       "      <td>3250.0</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>36.7</td>\n",
       "      <td>19.3</td>\n",
       "      <td>193.0</td>\n",
       "      <td>3450.0</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  species     island  culmen_length_mm  culmen_depth_mm  flipper_length_mm  \\\n",
       "0  Adelie  Torgersen              39.1             18.7              181.0   \n",
       "1  Adelie  Torgersen              39.5             17.4              186.0   \n",
       "2  Adelie  Torgersen              40.3             18.0              195.0   \n",
       "3  Adelie  Torgersen               NaN              NaN                NaN   \n",
       "4  Adelie  Torgersen              36.7             19.3              193.0   \n",
       "\n",
       "   body_mass_g     sex  \n",
       "0       3750.0    MALE  \n",
       "1       3800.0  FEMALE  \n",
       "2       3250.0  FEMALE  \n",
       "3          NaN     NaN  \n",
       "4       3450.0  FEMALE  "
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "%matplotlib inline\n",
    "\n",
    "data = pd.read_csv('penguins_size.csv')\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ec5a28fa-827c-442c-9506-dace9a2160ec",
   "metadata": {},
   "source": [
    "**Note: data for one Adelie species is missing (row 3)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e40e5dc0-bc67-4d6d-a2d3-4f2a818e5d29",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### Exploring Continuous features"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0188197b-3c3b-4077-b421-56dc9453c364",
   "metadata": {},
   "source": [
    "To explore continuous features we are going to drop the categorical features: `island` and `sex`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "3a981486-dc7e-4d3c-9672-393b2a9d93c9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>39.1</td>\n",
       "      <td>18.7</td>\n",
       "      <td>181.0</td>\n",
       "      <td>3750.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>39.5</td>\n",
       "      <td>17.4</td>\n",
       "      <td>186.0</td>\n",
       "      <td>3800.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>40.3</td>\n",
       "      <td>18.0</td>\n",
       "      <td>195.0</td>\n",
       "      <td>3250.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>36.7</td>\n",
       "      <td>19.3</td>\n",
       "      <td>193.0</td>\n",
       "      <td>3450.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  species  culmen_length_mm  culmen_depth_mm  flipper_length_mm  body_mass_g\n",
       "0  Adelie              39.1             18.7              181.0       3750.0\n",
       "1  Adelie              39.5             17.4              186.0       3800.0\n",
       "2  Adelie              40.3             18.0              195.0       3250.0\n",
       "3  Adelie               NaN              NaN                NaN          NaN\n",
       "4  Adelie              36.7             19.3              193.0       3450.0"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cat_feat = ['island', 'sex']\n",
    "data.drop(cat_feat, axis=1, inplace=True)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "ba945b4b-07a4-4525-81f3-242f9d03df94",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>342.000000</td>\n",
       "      <td>342.000000</td>\n",
       "      <td>342.000000</td>\n",
       "      <td>342.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>43.921930</td>\n",
       "      <td>17.151170</td>\n",
       "      <td>200.915205</td>\n",
       "      <td>4201.754386</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>5.459584</td>\n",
       "      <td>1.974793</td>\n",
       "      <td>14.061714</td>\n",
       "      <td>801.954536</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>32.100000</td>\n",
       "      <td>13.100000</td>\n",
       "      <td>172.000000</td>\n",
       "      <td>2700.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>39.225000</td>\n",
       "      <td>15.600000</td>\n",
       "      <td>190.000000</td>\n",
       "      <td>3550.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>44.450000</td>\n",
       "      <td>17.300000</td>\n",
       "      <td>197.000000</td>\n",
       "      <td>4050.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>48.500000</td>\n",
       "      <td>18.700000</td>\n",
       "      <td>213.000000</td>\n",
       "      <td>4750.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>59.600000</td>\n",
       "      <td>21.500000</td>\n",
       "      <td>231.000000</td>\n",
       "      <td>6300.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       culmen_length_mm  culmen_depth_mm  flipper_length_mm  body_mass_g\n",
       "count        342.000000       342.000000         342.000000   342.000000\n",
       "mean          43.921930        17.151170         200.915205  4201.754386\n",
       "std            5.459584         1.974793          14.061714   801.954536\n",
       "min           32.100000        13.100000         172.000000  2700.000000\n",
       "25%           39.225000        15.600000         190.000000  3550.000000\n",
       "50%           44.450000        17.300000         197.000000  4050.000000\n",
       "75%           48.500000        18.700000         213.000000  4750.000000\n",
       "max           59.600000        21.500000         231.000000  6300.000000"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "284f01e0-ec35-4a9b-bc51-a311144af7f0",
   "metadata": {},
   "source": [
    "based on the information on GitHub page (and the CSV file) the data set contains information on 344 penguins. Therefore, having the count 342 for each of continuous feature indicates that there are missing data for two data point (for 2 penguins)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "57cf6979-217c-4932-898a-4e8db0c24beb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>species</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Adelie</th>\n",
       "      <td>38.791391</td>\n",
       "      <td>18.346358</td>\n",
       "      <td>189.953642</td>\n",
       "      <td>3700.662252</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Chinstrap</th>\n",
       "      <td>48.833824</td>\n",
       "      <td>18.420588</td>\n",
       "      <td>195.823529</td>\n",
       "      <td>3733.088235</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Gentoo</th>\n",
       "      <td>47.504878</td>\n",
       "      <td>14.982114</td>\n",
       "      <td>217.186992</td>\n",
       "      <td>5076.016260</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           culmen_length_mm  culmen_depth_mm  flipper_length_mm  body_mass_g\n",
       "species                                                                     \n",
       "Adelie            38.791391        18.346358         189.953642  3700.662252\n",
       "Chinstrap         48.833824        18.420588         195.823529  3733.088235\n",
       "Gentoo            47.504878        14.982114         217.186992  5076.016260"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.groupby('species').mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d7b17419-4917-4d49-85a2-470723054aed",
   "metadata": {},
   "source": [
    "By grouping the data based on the species and calculating the mean, we notice few things:\n",
    "- Adelie species have shorter culmen length, wider culmen depth, shorter flippers and lighter body mass than  Gentoo.  \n",
    "- Adelie's only difference with Chinstrap is in their culmen length where Adelie has shorter culmen than Chinstrap.\n",
    "- On the other hand the only similarity between Gentoo and Chinstrap is in their culmen length and the two differ in every other feature. \n",
    "\n",
    "One strategy to differentiate between three species would be to first group the given example/penguin in either Adelie or Chinstrap+Gentoo based on its `culmen_length` then distinguish between Chinstrap and Gentoo species based on either `culmen_depth`, `flipper_length`, or `body_mass`. \n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bd465ab7-019b-4ccd-bb73-155441e2ed48",
   "metadata": {
    "tags": []
   },
   "source": [
    "#### Plotting continuous features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "db18b514-d0ad-477a-b45e-4ed44118d51c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'culmen lenght (mm) by species')"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#plotting culmen lenght for each species\n",
    "plt.figure(figsize=(8,5))\n",
    "sns.boxplot(x='species',y='culmen_length_mm',data=data, palette='rainbow')\n",
    "plt.title(\"culmen lenght (mm) by species\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7942bbcf-623e-4697-9aae-92f0f33be9f3",
   "metadata": {},
   "source": [
    "box-plots have about same length suggesting that the culmen length for each species is evenly distributed. we can see that the Adelie species are fairly separated from the other two species and the data is normal. \n",
    "Two things to note are:\n",
    "1. data for the chinstrap species is slightly skewed to left \n",
    "2. there is one outlier present in the Gentoo dataset. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "2188c0ac-8264-4a06-93e6-a77482c9dfd6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'culmen depth (mm) by species')"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#plotting culmen depth for each species\n",
    "plt.figure(figsize=(8,5))\n",
    "sns.boxplot(x='species',y='culmen_depth_mm',data=data, palette='rainbow')\n",
    "plt.title(\"culmen depth (mm) by species\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8bba8722-45af-4c2d-9ec1-d80be08b2a9a",
   "metadata": {},
   "source": [
    "We can see that the Chinstrap species has slightly wider range when it comes to culmen depth than other two species. other than that the data is fairly distributed. The Gentoo species is reasonably separated from Adelie and Chinstrap which have a very close median. Final thing to note is the presence of one outlier in the Adelie dataset. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "d518399e-20fa-4d2d-b265-615e7cd45a37",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Flipper length (mm) by species')"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#plotting flipper length for each species\n",
    "plt.figure(figsize=(8,5))\n",
    "sns.boxplot(x='species',y='flipper_length_mm',data=data, palette='rainbow')\n",
    "plt.title(\"Flipper length (mm) by species\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5d1b2daf-81fa-456f-ae5e-3ac768b57322",
   "metadata": {},
   "source": [
    "The Gentoo species is separated from other two species in this feature like in culmen_depth. Adelie and Chinstrap again have a fairly close median to each other with Adelie having two outlier in either high and low end of the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "5d97ab89-79f3-48f1-8482-82e8e3a47323",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Body Mass(g) by species')"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#plotting body mass for each species\n",
    "plt.figure(figsize=(8,5))\n",
    "sns.boxplot(x='species',y='body_mass_g',data=data, palette='rainbow')\n",
    "plt.title(\"Body Mass(g) by species\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fd018701-2179-4ace-841c-f70dc0fa9df0",
   "metadata": {},
   "source": [
    "As for the body mass, Gentoo species is one more time distinguished from the other two. Adelie and Chinstrap species have similar median with the difference that the Chinstrap species' body mass is more closely concentrated around the median than Adelie's. One final thing that the graph tells is the presence of two outlier in high and low ends of Chinstrap data. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "839ec4a5-7543-4bcb-aade-4a15318572a7",
   "metadata": {},
   "source": [
    "In conclusion, except few outliers our data set is fairly distributed evenly. We can use the culmen length to differentiate between Adelie and the other two species. to differentiate between Chinstrap and Gentoo we can use either `culmen_depth`, `flipper_length`, or `body_mass` (if the computation is expensive or use all three if the computation is possible)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "44c94c15-6bc2-41ce-8354-c5ee4a464c87",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>39.1</td>\n",
       "      <td>18.7</td>\n",
       "      <td>181.0</td>\n",
       "      <td>3750.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>39.5</td>\n",
       "      <td>17.4</td>\n",
       "      <td>186.0</td>\n",
       "      <td>3800.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>40.3</td>\n",
       "      <td>18.0</td>\n",
       "      <td>195.0</td>\n",
       "      <td>3250.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>36.7</td>\n",
       "      <td>19.3</td>\n",
       "      <td>193.0</td>\n",
       "      <td>3450.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>39.3</td>\n",
       "      <td>20.6</td>\n",
       "      <td>190.0</td>\n",
       "      <td>3650.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>339</th>\n",
       "      <td>Gentoo</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    species  culmen_length_mm  culmen_depth_mm  flipper_length_mm  body_mass_g\n",
       "0    Adelie              39.1             18.7              181.0       3750.0\n",
       "1    Adelie              39.5             17.4              186.0       3800.0\n",
       "2    Adelie              40.3             18.0              195.0       3250.0\n",
       "3    Adelie               NaN              NaN                NaN          NaN\n",
       "4    Adelie              36.7             19.3              193.0       3450.0\n",
       "5    Adelie              39.3             20.6              190.0       3650.0\n",
       "339  Gentoo               NaN              NaN                NaN          NaN"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.groupby(data['culmen_length_mm'].isnull()).head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "56fa2af8-b274-4627-8156-f24ded69cc90",
   "metadata": {},
   "source": [
    "by grouping the data using culmen_length to find the two missing data, we notice that one missing data was for Adelie and the other was for Gentoo species. In both cases all the continuous features were missing justifying the missing values for the other features. \n",
    "There is no indication that missing values are correlated with specific continuous feature or the target variable so we can conclude that these values are missing at random. In the next steps we fill the missing values with average of their corresponding feature. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "c93b83f0-45fb-4296-8857-bbfd9d316ddd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "species              0\n",
       "culmen_length_mm     0\n",
       "culmen_depth_mm      0\n",
       "flipper_length_mm    0\n",
       "body_mass_g          0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['culmen_length_mm'].fillna(data['culmen_length_mm'].mean(), inplace=True) \n",
    "data['culmen_depth_mm'].fillna(data['culmen_depth_mm'].mean(), inplace=True) \n",
    "data['flipper_length_mm'].fillna(data['flipper_length_mm'].mean(), inplace=True) \n",
    "data['body_mass_g'].fillna(data['body_mass_g'].mean(), inplace=True) \n",
    "data.isnull().sum() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "bb850d53-b1ac-4976-a7a4-2c322b68c050",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>39.10000</td>\n",
       "      <td>18.70000</td>\n",
       "      <td>181.000000</td>\n",
       "      <td>3750.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>39.50000</td>\n",
       "      <td>17.40000</td>\n",
       "      <td>186.000000</td>\n",
       "      <td>3800.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>40.30000</td>\n",
       "      <td>18.00000</td>\n",
       "      <td>195.000000</td>\n",
       "      <td>3250.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>43.92193</td>\n",
       "      <td>17.15117</td>\n",
       "      <td>200.915205</td>\n",
       "      <td>4201.754386</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>36.70000</td>\n",
       "      <td>19.30000</td>\n",
       "      <td>193.000000</td>\n",
       "      <td>3450.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  species  culmen_length_mm  culmen_depth_mm  flipper_length_mm  body_mass_g\n",
       "0  Adelie          39.10000         18.70000         181.000000  3750.000000\n",
       "1  Adelie          39.50000         17.40000         186.000000  3800.000000\n",
       "2  Adelie          40.30000         18.00000         195.000000  3250.000000\n",
       "3  Adelie          43.92193         17.15117         200.915205  4201.754386\n",
       "4  Adelie          36.70000         19.30000         193.000000  3450.000000"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c2fd634b-93fe-4599-a5e3-e5c9a53cda92",
   "metadata": {},
   "source": [
    "We see that one of the rows of missing data (row 3) is replaced with the average values of each feature. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ec594e56-1154-4ebe-b59b-52437a04ef6a",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### Exploring Categorical features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "303190f4-024f-45fd-bf7f-1fb51bce4bde",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>island</th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "      <th>sex</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>39.1</td>\n",
       "      <td>18.7</td>\n",
       "      <td>181.0</td>\n",
       "      <td>3750.0</td>\n",
       "      <td>MALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>39.5</td>\n",
       "      <td>17.4</td>\n",
       "      <td>186.0</td>\n",
       "      <td>3800.0</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>40.3</td>\n",
       "      <td>18.0</td>\n",
       "      <td>195.0</td>\n",
       "      <td>3250.0</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>36.7</td>\n",
       "      <td>19.3</td>\n",
       "      <td>193.0</td>\n",
       "      <td>3450.0</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  species     island  culmen_length_mm  culmen_depth_mm  flipper_length_mm  \\\n",
       "0  Adelie  Torgersen              39.1             18.7              181.0   \n",
       "1  Adelie  Torgersen              39.5             17.4              186.0   \n",
       "2  Adelie  Torgersen              40.3             18.0              195.0   \n",
       "3  Adelie  Torgersen               NaN              NaN                NaN   \n",
       "4  Adelie  Torgersen              36.7             19.3              193.0   \n",
       "\n",
       "   body_mass_g     sex  \n",
       "0       3750.0    MALE  \n",
       "1       3800.0  FEMALE  \n",
       "2       3250.0  FEMALE  \n",
       "3          NaN     NaN  \n",
       "4       3450.0  FEMALE  "
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = pd.read_csv('penguins_size.csv')\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "82931fa7-cdd2-4b68-ad17-84263accf12a",
   "metadata": {},
   "source": [
    "to explore categorical features, first we drop all the continuous features: `culmen_length_mm`, `culmen_depth_mm`, `flipper_length_mm`, `body_mass_g`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "6822acb6-2da2-467c-b494-534f62c4db3b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>island</th>\n",
       "      <th>sex</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>MALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  species     island     sex\n",
       "0  Adelie  Torgersen    MALE\n",
       "1  Adelie  Torgersen  FEMALE\n",
       "2  Adelie  Torgersen  FEMALE\n",
       "3  Adelie  Torgersen     NaN\n",
       "4  Adelie  Torgersen  FEMALE"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "con_feat = ['culmen_length_mm', 'culmen_depth_mm', 'flipper_length_mm', 'body_mass_g']\n",
    "data.drop(con_feat, axis=1, inplace=True )\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "4f5083d5-b6a6-46a9-a801-6166f0ff6bf7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 344 entries, 0 to 343\n",
      "Data columns (total 3 columns):\n",
      " #   Column   Non-Null Count  Dtype \n",
      "---  ------   --------------  ----- \n",
      " 0   species  344 non-null    object\n",
      " 1   island   344 non-null    object\n",
      " 2   sex      334 non-null    object\n",
      "dtypes: object(3)\n",
      "memory usage: 8.2+ KB\n"
     ]
    }
   ],
   "source": [
    "data.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0a527cd2-8669-48c3-b6ce-f7c95eb40882",
   "metadata": {},
   "source": [
    "By using info() method we can see that all the features are objects. There are no missing values for the `island` feature but 10 values are missing in the `sex` feature. In the next step, we have to see whether this values are missing randomly or not:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "6d736710-65a9-4ee7-bd5f-feeffa965469",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>island</th>\n",
       "      <th>sex</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>MALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>MALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>MALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>MALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>MALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Dream</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>246</th>\n",
       "      <td>Gentoo</td>\n",
       "      <td>Biscoe</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>286</th>\n",
       "      <td>Gentoo</td>\n",
       "      <td>Biscoe</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>324</th>\n",
       "      <td>Gentoo</td>\n",
       "      <td>Biscoe</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>339</th>\n",
       "      <td>Gentoo</td>\n",
       "      <td>Biscoe</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    species     island     sex\n",
       "0    Adelie  Torgersen    MALE\n",
       "1    Adelie  Torgersen  FEMALE\n",
       "2    Adelie  Torgersen  FEMALE\n",
       "3    Adelie  Torgersen     NaN\n",
       "4    Adelie  Torgersen  FEMALE\n",
       "5    Adelie  Torgersen    MALE\n",
       "6    Adelie  Torgersen  FEMALE\n",
       "7    Adelie  Torgersen    MALE\n",
       "8    Adelie  Torgersen     NaN\n",
       "9    Adelie  Torgersen     NaN\n",
       "10   Adelie  Torgersen     NaN\n",
       "11   Adelie  Torgersen     NaN\n",
       "12   Adelie  Torgersen  FEMALE\n",
       "13   Adelie  Torgersen    MALE\n",
       "14   Adelie  Torgersen    MALE\n",
       "47   Adelie      Dream     NaN\n",
       "246  Gentoo     Biscoe     NaN\n",
       "286  Gentoo     Biscoe     NaN\n",
       "324  Gentoo     Biscoe     NaN\n",
       "339  Gentoo     Biscoe     NaN"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.groupby(data['sex'].isnull()).head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "25e849d3-809a-478e-ad67-7fead1841422",
   "metadata": {},
   "source": [
    "We can see that most of missing values for `sex` feature are for Adelie species in Torgersen island(5). There are also values missing for Adelie in other islands and some missing values for Gentoo as well suggesting that these missing values are mostly random. \n",
    "\n",
    "Since `sex` feature does not affect the species of a penguin, this missing values are irrelevant and we can ignore the `sex` feature all together and only focus on `island`.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fd0b819a-41b5-4cda-a684-5c21f70c5257",
   "metadata": {},
   "source": [
    "#### Plotting categorical features"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "81844da9-dc93-427e-b253-0dcbe78bd417",
   "metadata": {},
   "source": [
    "We are going to use count plot of each species in each island to see how they are distributed based on their recorded island:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "1c6e2e78-725b-4c4e-8218-8f80e46f0bf2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Count of each species in each island')"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8,5))\n",
    "sns.countplot(x='island',data=data, palette='rainbow',hue='species')\n",
    "plt.title(\"Count of each species in each island\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "70e78874-b1dc-4bce-a26a-df0c32a2b5f8",
   "metadata": {},
   "source": [
    "Based on the count-plot above we can see that Adelie species are found in any of the 3 islands while Gentoo are only found in Biscoe and Chinstrap are only found in Dream island. By using pivot table, we can further support this finding:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "2031ce9b-100e-4a26-a5a7-3004142af5ec",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>island</th>\n",
       "      <th>Biscoe</th>\n",
       "      <th>Dream</th>\n",
       "      <th>Torgersen</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>species</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Adelie</th>\n",
       "      <td>44.0</td>\n",
       "      <td>55.0</td>\n",
       "      <td>47.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Chinstrap</th>\n",
       "      <td>NaN</td>\n",
       "      <td>68.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Gentoo</th>\n",
       "      <td>120.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "island     Biscoe  Dream  Torgersen\n",
       "species                            \n",
       "Adelie       44.0   55.0       47.0\n",
       "Chinstrap     NaN   68.0        NaN\n",
       "Gentoo      120.0    NaN        NaN"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.pivot_table('species', index='species', columns='island', aggfunc='count')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5d291fb6-9d42-42f8-bdd8-80c979b38b7d",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "## **Part II: Cleaning data**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "da3cac97-47f1-4fbd-9ef9-a5e801bafa87",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>island</th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "      <th>sex</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>39.1</td>\n",
       "      <td>18.7</td>\n",
       "      <td>181.0</td>\n",
       "      <td>3750.0</td>\n",
       "      <td>MALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>39.5</td>\n",
       "      <td>17.4</td>\n",
       "      <td>186.0</td>\n",
       "      <td>3800.0</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>40.3</td>\n",
       "      <td>18.0</td>\n",
       "      <td>195.0</td>\n",
       "      <td>3250.0</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>36.7</td>\n",
       "      <td>19.3</td>\n",
       "      <td>193.0</td>\n",
       "      <td>3450.0</td>\n",
       "      <td>FEMALE</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  species     island  culmen_length_mm  culmen_depth_mm  flipper_length_mm  \\\n",
       "0  Adelie  Torgersen              39.1             18.7              181.0   \n",
       "1  Adelie  Torgersen              39.5             17.4              186.0   \n",
       "2  Adelie  Torgersen              40.3             18.0              195.0   \n",
       "3  Adelie  Torgersen               NaN              NaN                NaN   \n",
       "4  Adelie  Torgersen              36.7             19.3              193.0   \n",
       "\n",
       "   body_mass_g     sex  \n",
       "0       3750.0    MALE  \n",
       "1       3800.0  FEMALE  \n",
       "2       3250.0  FEMALE  \n",
       "3          NaN     NaN  \n",
       "4       3450.0  FEMALE  "
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = pd.read_csv('penguins_size.csv')\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6bb1f5b6-1255-424c-8e98-feaec8d80f69",
   "metadata": {},
   "source": [
    "To clean the data:\n",
    "1. First we remove the `sex` feature since the sex of penguin does not determine its species.\n",
    "2. We fill the missing values with the average values of their respective feature\n",
    "3. we convert `species` to numeric to ease the model computation :\n",
    "    - Adelie --> 0\n",
    "    - Chinstrap --> 1\n",
    "    - Gentoo --> 2\n",
    "4. we convert `island` to numeric for the same reason:\n",
    "    - Dream --> 0\n",
    "    - Torgersen --> 1\n",
    "    - Biscoe --> 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "4817fd3d-fc85-40a2-91aa-4be87f55d2ee",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>island</th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>39.1</td>\n",
       "      <td>18.7</td>\n",
       "      <td>181.0</td>\n",
       "      <td>3750.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>39.5</td>\n",
       "      <td>17.4</td>\n",
       "      <td>186.0</td>\n",
       "      <td>3800.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>40.3</td>\n",
       "      <td>18.0</td>\n",
       "      <td>195.0</td>\n",
       "      <td>3250.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Adelie</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>36.7</td>\n",
       "      <td>19.3</td>\n",
       "      <td>193.0</td>\n",
       "      <td>3450.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  species     island  culmen_length_mm  culmen_depth_mm  flipper_length_mm  \\\n",
       "0  Adelie  Torgersen              39.1             18.7              181.0   \n",
       "1  Adelie  Torgersen              39.5             17.4              186.0   \n",
       "2  Adelie  Torgersen              40.3             18.0              195.0   \n",
       "3  Adelie  Torgersen               NaN              NaN                NaN   \n",
       "4  Adelie  Torgersen              36.7             19.3              193.0   \n",
       "\n",
       "   body_mass_g  \n",
       "0       3750.0  \n",
       "1       3800.0  \n",
       "2       3250.0  \n",
       "3          NaN  \n",
       "4       3450.0  "
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#step1: removing the feature `sex`\n",
    "data.drop(['sex'], axis=1, inplace=True)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "1f1d58e5-313d-427a-ba34-6eee5c8c306a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "species              0\n",
       "island               0\n",
       "culmen_length_mm     2\n",
       "culmen_depth_mm      2\n",
       "flipper_length_mm    2\n",
       "body_mass_g          2\n",
       "dtype: int64"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "422550a1-605f-4e4f-b450-345996ec9fd0",
   "metadata": {},
   "outputs": [],
   "source": [
    "#step 2: filling missing values\n",
    "\n",
    "data['culmen_length_mm'].fillna(data['culmen_length_mm'].mean(), inplace=True)\n",
    "data['culmen_depth_mm'].fillna(data['culmen_depth_mm'].mean(), inplace=True)\n",
    "data['flipper_length_mm'].fillna(data['flipper_length_mm'].mean(), inplace=True)\n",
    "data['body_mass_g'].fillna(data['body_mass_g'].mean(), inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "0ded71d0-6756-463a-ad28-9e1da8f547f6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "species              0\n",
       "island               0\n",
       "culmen_length_mm     0\n",
       "culmen_depth_mm      0\n",
       "flipper_length_mm    0\n",
       "body_mass_g          0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "dbfbc56c-f8ec-4efd-80d5-6525588d632e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>island</th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>39.10000</td>\n",
       "      <td>18.70000</td>\n",
       "      <td>181.000000</td>\n",
       "      <td>3750.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>39.50000</td>\n",
       "      <td>17.40000</td>\n",
       "      <td>186.000000</td>\n",
       "      <td>3800.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>40.30000</td>\n",
       "      <td>18.00000</td>\n",
       "      <td>195.000000</td>\n",
       "      <td>3250.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>43.92193</td>\n",
       "      <td>17.15117</td>\n",
       "      <td>200.915205</td>\n",
       "      <td>4201.754386</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>Torgersen</td>\n",
       "      <td>36.70000</td>\n",
       "      <td>19.30000</td>\n",
       "      <td>193.000000</td>\n",
       "      <td>3450.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   species     island  culmen_length_mm  culmen_depth_mm  flipper_length_mm  \\\n",
       "0        0  Torgersen          39.10000         18.70000         181.000000   \n",
       "1        0  Torgersen          39.50000         17.40000         186.000000   \n",
       "2        0  Torgersen          40.30000         18.00000         195.000000   \n",
       "3        0  Torgersen          43.92193         17.15117         200.915205   \n",
       "4        0  Torgersen          36.70000         19.30000         193.000000   \n",
       "\n",
       "   body_mass_g  \n",
       "0  3750.000000  \n",
       "1  3800.000000  \n",
       "2  3250.000000  \n",
       "3  4201.754386  \n",
       "4  3450.000000  "
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#step3: converting species to numeric\n",
    "#Adelie --> 0\n",
    "#Chinstrap --> 1\n",
    "#Gentoo --> 2\n",
    "\n",
    "spe_num = {'Adelie': 0, 'Chinstrap': 1, 'Gentoo': 2}\n",
    "\n",
    "data['species'] = data['species'].map(spe_num)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "08a0d563-e7e0-46ec-948b-eebafb409b60",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>island</th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>39.10000</td>\n",
       "      <td>18.70000</td>\n",
       "      <td>181.000000</td>\n",
       "      <td>3750.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>39.50000</td>\n",
       "      <td>17.40000</td>\n",
       "      <td>186.000000</td>\n",
       "      <td>3800.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>40.30000</td>\n",
       "      <td>18.00000</td>\n",
       "      <td>195.000000</td>\n",
       "      <td>3250.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>43.92193</td>\n",
       "      <td>17.15117</td>\n",
       "      <td>200.915205</td>\n",
       "      <td>4201.754386</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>36.70000</td>\n",
       "      <td>19.30000</td>\n",
       "      <td>193.000000</td>\n",
       "      <td>3450.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   species  island  culmen_length_mm  culmen_depth_mm  flipper_length_mm  \\\n",
       "0        0       1          39.10000         18.70000         181.000000   \n",
       "1        0       1          39.50000         17.40000         186.000000   \n",
       "2        0       1          40.30000         18.00000         195.000000   \n",
       "3        0       1          43.92193         17.15117         200.915205   \n",
       "4        0       1          36.70000         19.30000         193.000000   \n",
       "\n",
       "   body_mass_g  \n",
       "0  3750.000000  \n",
       "1  3800.000000  \n",
       "2  3250.000000  \n",
       "3  4201.754386  \n",
       "4  3450.000000  "
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# step 4: converting islands to numeric\n",
    "#Dream --> 0\n",
    "#Torgersen --> 1\n",
    "#Biscoe --> 2\n",
    "\n",
    "isl_num = {'Dream': 0, 'Torgersen': 1, 'Biscoe': 2}\n",
    "\n",
    "data['island'] = data['island'].map(isl_num)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d1a6371f-343e-47cd-9d56-88e5951b625a",
   "metadata": {},
   "source": [
    "After cleaning we save the cleaned data as follows:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "f800463b-dd93-4704-816d-99d4feaa0dbe",
   "metadata": {},
   "outputs": [],
   "source": [
    "data.to_csv('penguins_cleaned.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9b541dcf-7006-4676-95c9-7838b11dab1a",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "## **Part III: Splitting data 60-20-20**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7e90a88f-e07d-40bb-922a-494a52db2e16",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### Reading cleaned data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "8a58986c-d04c-45de-999c-564b763ce08e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>island</th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>39.10000</td>\n",
       "      <td>18.70000</td>\n",
       "      <td>181.000000</td>\n",
       "      <td>3750.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>39.50000</td>\n",
       "      <td>17.40000</td>\n",
       "      <td>186.000000</td>\n",
       "      <td>3800.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>40.30000</td>\n",
       "      <td>18.00000</td>\n",
       "      <td>195.000000</td>\n",
       "      <td>3250.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>43.92193</td>\n",
       "      <td>17.15117</td>\n",
       "      <td>200.915205</td>\n",
       "      <td>4201.754386</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>36.70000</td>\n",
       "      <td>19.30000</td>\n",
       "      <td>193.000000</td>\n",
       "      <td>3450.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   species  island  culmen_length_mm  culmen_depth_mm  flipper_length_mm  \\\n",
       "0        0       1          39.10000         18.70000         181.000000   \n",
       "1        0       1          39.50000         17.40000         186.000000   \n",
       "2        0       1          40.30000         18.00000         195.000000   \n",
       "3        0       1          43.92193         17.15117         200.915205   \n",
       "4        0       1          36.70000         19.30000         193.000000   \n",
       "\n",
       "   body_mass_g  \n",
       "0  3750.000000  \n",
       "1  3800.000000  \n",
       "2  3250.000000  \n",
       "3  4201.754386  \n",
       "4  3450.000000  "
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "data = pd.read_csv('penguins_cleaned.csv')\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a1f0e37b-13e2-4438-8c59-f51a2d3d9d29",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### Splitting data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "41d9823b-7dfe-488d-9a5f-cca6ac946df9",
   "metadata": {},
   "outputs": [],
   "source": [
    "features = data.drop('species', axis=1)\n",
    "labels = data['species']\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(features, labels, test_size=0.4, random_state=42)\n",
    "X_test, X_val, y_test, y_val = train_test_split(X_test, y_test, test_size=0.5, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "54d2b402-528f-488d-8b6b-3c93f08e52b8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.6\n",
      "0.2\n",
      "0.2\n"
     ]
    }
   ],
   "source": [
    "for dataset in [y_train, y_val, y_test]:\n",
    "    print(round(len(dataset) / len(labels), 2))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e640743d-1778-4a60-831d-157aa8ddbed9",
   "metadata": {},
   "source": [
    "Saving the three datasets:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "c21eb3eb-088c-4c2d-b2c8-674ef552a88c",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train.to_csv('train_features.csv', index=False)\n",
    "X_val.to_csv('val_features.csv', index=False)\n",
    "X_test.to_csv('test_features.csv', index=False)\n",
    "\n",
    "y_train.to_csv('train_labels.csv', index=False)\n",
    "y_val.to_csv('val_labels.csv', index=False)\n",
    "y_test.to_csv('test_labels.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9cf52929-7855-4cfc-a5d9-4c48f8207c62",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### **Logistic regression algorithm:**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "1688acf6-cc33-46a4-9547-d7c58e030e53",
   "metadata": {},
   "outputs": [],
   "source": [
    "import joblib\n",
    "import pandas as pd\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.model_selection import GridSearchCV # used for tuning algorithms and hyperparameters\n",
    "from sklearn.svm import SVC\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore', category=FutureWarning)\n",
    "warnings.filterwarnings('ignore', category=DeprecationWarning)\n",
    "\n",
    "tr_features = pd.read_csv('train_features.csv')\n",
    "tr_labels = pd.read_csv('train_labels.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "260a46bd-76a1-422b-8789-e164c67cf406",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Hyperparameter tuning: C"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "f0281fc0-e1ec-4966-9704-dcd6b06b7bb0",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Citation 2\n",
    "def print_results(results):\n",
    "    print('BEST PARAMS: {}\\n'.format(results.best_params_))\n",
    "\n",
    "    means = results.cv_results_['mean_test_score']\n",
    "    stds = results.cv_results_['std_test_score']\n",
    "    for mean, std, params in zip(means, stds, results.cv_results_['params']):\n",
    "        print('{} (+/-{}) for {}'.format(round(mean, 3), round(std * 2, 3), params))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "daf34ce8-2c3e-44c1-b4bf-b997c387ee5c",
   "metadata": {},
   "source": [
    "** because of the convergence warning, increased the max_iter to 3000. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "21c5b303-9b48-4300-b938-f2b8528e6737",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BEST PARAMS: {'C': 100}\n",
      "\n",
      "0.893 (+/-0.079) for {'C': 0.001}\n",
      "0.971 (+/-0.036) for {'C': 0.01}\n",
      "0.981 (+/-0.019) for {'C': 0.1}\n",
      "0.981 (+/-0.019) for {'C': 1}\n",
      "0.985 (+/-0.024) for {'C': 10}\n",
      "0.995 (+/-0.02) for {'C': 100}\n",
      "0.985 (+/-0.039) for {'C': 1000}\n"
     ]
    }
   ],
   "source": [
    "lr = LogisticRegression(max_iter=3000)\n",
    "parameters = {\n",
    "    'C': [0.001, 0.01, 0.1, 1, 10, 100, 1000]\n",
    "}\n",
    "\n",
    "cv = GridSearchCV(lr, parameters, cv=5)\n",
    "cv.fit(tr_features, tr_labels.values.ravel())\n",
    "\n",
    "print_results(cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7ab8782d-6d42-4917-b798-86f14dba3d99",
   "metadata": {},
   "source": [
    "With the max_iter of 3000, the best C with highest performance is C=100 which has result of 99.5% suggesting that C = 100 allows the model to differentiate between the three species. \n",
    "A\n",
    "s C decreases the model performance also decreases resulting in under-fitting and as C increase after 100, the performance also slightly drops resulting in possible over-fitting. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "4e1db70a-5992-4904-a857-01544303a8b5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=100, max_iter=3000)"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "88d93e2d-cb7f-4010-a3ae-dc45510505e8",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Learning Curve:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "abd7dfef-e318-41fd-86d9-e4553b37d99c",
   "metadata": {},
   "outputs": [],
   "source": [
    "#citation 3\n",
    "from sklearn.model_selection import learning_curve\n",
    "# the training data is 60% of the whole data --> 206\n",
    "# we are doing 5-fold cross validation with 80% training set and 20% validation set:\n",
    "# 206 training example : 80% training (164) 20% validation (42) in cross validation\n",
    "train_sizes = [5, 20, 40, 60, 80, 100, 120, 140, 164]\n",
    "train_sizes, train_scores, validation_scores = learning_curve(\n",
    "estimator = LogisticRegression(C=100, max_iter=3000),\n",
    "X = tr_features,\n",
    "y = tr_labels.values.ravel(), train_sizes = train_sizes, cv = 5,\n",
    "scoring = 'neg_mean_squared_error')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "2ffbc405-dbb1-42f1-bc3f-6bf7f6f5e2cf",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean training scores\n",
      "\n",
      " 5     -0.0\n",
      "20    -0.0\n",
      "40    -0.0\n",
      "60    -0.0\n",
      "80    -0.0\n",
      "100   -0.0\n",
      "120   -0.0\n",
      "140   -0.0\n",
      "164   -0.0\n",
      "dtype: float64\n",
      "\n",
      " --------------------\n",
      "\n",
      "Mean validation scores\n",
      "\n",
      " 5      0.274100\n",
      "20     0.072242\n",
      "40     0.009756\n",
      "60     0.009756\n",
      "80    -0.000000\n",
      "100    0.004762\n",
      "120    0.004762\n",
      "140    0.004878\n",
      "164    0.009756\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "train_scores_mean = -train_scores.mean(axis = 1)\n",
    "validation_scores_mean = -validation_scores.mean(axis = 1)\n",
    "print('Mean training scores\\n\\n', pd.Series(train_scores_mean, index = train_sizes))\n",
    "print('\\n', '-' * 20) # separator\n",
    "print('\\nMean validation scores\\n\\n',pd.Series(validation_scores_mean, index = train_sizes))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "82af531c-9370-48b4-923e-09943a9a6bfb",
   "metadata": {},
   "source": [
    "when the model is in training, we can see that it causes 0 error regardless of training set size. This suggest the LR model is able to fit and predict perfectly in training mode. When model is tested using the validation set we can see that when the size is 5 we have the highest error score. This is because the model was trained on the a small number of data point so it is unable to make generalization and made lots of errors when tested with the validation set. As the size of the training set increases, the error rate decreases and it reaches 0 when the size is 80 and then increases again. Error scores not remaining constant suggest that having more data points for training possibly would improve the performance of LR model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "c76e0bae-e938-4b4e-a9cc-704c9ba3e09a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 0.3)"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x396 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.style.use('seaborn')\n",
    "plt.plot(train_sizes, train_scores_mean, label = 'Training error')\n",
    "plt.plot(train_sizes, validation_scores_mean, label = 'Validation error')\n",
    "plt.ylabel('MSE', fontsize = 14)\n",
    "plt.xlabel('Training set size', fontsize = 14)\n",
    "plt.title('Learning curves for a LR model', fontsize = 18, y = 1.03)\n",
    "plt.legend()\n",
    "plt.ylim(0,0.3)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "87632e79-f064-479a-b8c1-0ac7fbb5d703",
   "metadata": {},
   "source": [
    "One thing of the note is that the training score for all the different training sizes is 0. This either means that the model is able to fit and predict perfectly while in training mode or there was some sort of human error in the process of outputting the scores.  Based on a quick research I have done the training score should not be zero and as the training set size increases the number of error should also increase (even if slightly). Since I could not find the error that might have set the training error scores to zero, I will assume that this model is fitting the data perfectly. \n",
    "\n",
    "The learning curve further supports the previous discussion an we can see that  as the size of the training set increases the number of errors that model makes when making predictions for the validation set decreases, suggesting that model is improving at generalizing the data for the validation examples. The number of error in validation mode reaches 0 when the training size is around 80 and starts increasing slightly afterwards. Assuming that the training error is correct and since the two training and validation lines do not converge, increasing data points for training might be useful. \n",
    "\n",
    "Since both training and validation errors are low and the gap between the two is narrow, we can conclude that as the training set size increases, we have a model with lower bias and variance meaning our model has the optimum complexity. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "26870aea-e9a2-40e3-9f7f-96cfedcde369",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['LR_model.pkl']"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "joblib.dump(cv.best_estimator_, 'LR_model.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1ff55ae2-85be-49c4-b17b-ccc9b65c9b9b",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### **Support Vector Machine (SVM)**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b9a8de07-346e-41cc-ae4c-d222316deb67",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Hyperparameter Tuning: C & Kernel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "b32adab7-527c-4970-aa1d-5d4a40716c8a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BEST PARAMS: {'C': 0.1, 'kernel': 'linear'}\n",
      "\n",
      "0.99 (+/-0.024) for {'C': 0.1, 'kernel': 'linear'}\n",
      "0.748 (+/-0.021) for {'C': 0.1, 'kernel': 'rbf'}\n",
      "0.985 (+/-0.039) for {'C': 1, 'kernel': 'linear'}\n",
      "0.748 (+/-0.048) for {'C': 1, 'kernel': 'rbf'}\n",
      "0.99 (+/-0.039) for {'C': 10, 'kernel': 'linear'}\n",
      "0.743 (+/-0.049) for {'C': 10, 'kernel': 'rbf'}\n"
     ]
    }
   ],
   "source": [
    "svc = SVC()\n",
    "parameters = { # used to search for the hyperparameters\n",
    "    'kernel': ['linear', 'rbf'],\n",
    "    'C': [0.1, 1, 10]\n",
    "}\n",
    "\n",
    "cv = GridSearchCV(svc, parameters, cv=5) # cv = 5 --> 5 fold validation\n",
    "cv.fit(tr_features, tr_labels.values.ravel())\n",
    "\n",
    "print_results(cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6b6b95b2-a257-4f7a-b42d-f716bdce7553",
   "metadata": {},
   "source": [
    "Based on the above values, we can see that kernel has a great impact on the performance of the model and linear kernel is doing relatively better than rbf suggesting that our data is linearly separable. We can also see that the performance of the three values of C are fairly similar with minor difference but overall the linear kernel with C = 0.1 has the best performance with 99% performance rate. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "e834e639-d87b-4564-a7e1-fca9f118123e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SVC(C=0.1, kernel='linear')"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f2a83b5a-c0af-4618-ac3f-4d50bff678bb",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Learning Curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "46bc5182-de22-4ffc-aefc-4105fc3a0e24",
   "metadata": {},
   "outputs": [],
   "source": [
    "#citation 3\n",
    "from sklearn.model_selection import learning_curve\n",
    "# the training data is 60% of the whole data --> 206\n",
    "# we are doing 5-fold cross validation with 80% training set and 20% validation set:\n",
    "# 206 training example : 80% training (164) 20% validation (42) in cross validation\n",
    "train_sizes = [5, 20, 40, 60, 80, 100, 120, 140, 164]\n",
    "train_sizes, train_scores, validation_scores = learning_curve(\n",
    "estimator = SVC(C=0.1, kernel='linear'),\n",
    "X = tr_features,\n",
    "y = tr_labels.values.ravel(), train_sizes = train_sizes, cv = 5,\n",
    "scoring = 'neg_mean_squared_error')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "5584efd6-8f33-408c-b633-0a60af2c92e0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean training scores\n",
      "\n",
      " 5     -0.000000\n",
      "20     0.040000\n",
      "40     0.025000\n",
      "60     0.020000\n",
      "80     0.015000\n",
      "100    0.004000\n",
      "120    0.003333\n",
      "140    0.004286\n",
      "164    0.003659\n",
      "dtype: float64\n",
      "\n",
      " --------------------\n",
      "\n",
      "Mean validation scores\n",
      "\n",
      " 5      0.343670\n",
      "20     0.125784\n",
      "40     0.082578\n",
      "60     0.063066\n",
      "80     0.048664\n",
      "100    0.058188\n",
      "120    0.019164\n",
      "140    0.009640\n",
      "164    0.009640\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "train_scores_mean = -train_scores.mean(axis = 1)\n",
    "validation_scores_mean = -validation_scores.mean(axis = 1)\n",
    "print('Mean training scores\\n\\n', pd.Series(train_scores_mean, index = train_sizes))\n",
    "print('\\n', '-' * 20) # separator\n",
    "print('\\nMean validation scores\\n\\n',pd.Series(validation_scores_mean, index = train_sizes))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ac35817f-ca11-480e-9e43-d41a925297d0",
   "metadata": {},
   "source": [
    "When the  model is training with very small size of the training set (5) we can see that it makes 0 errors and as the the training set size increases the errors model make also changes and starts to decrease. When the model is being tested on the validation set we can see that for small size of the training set it makes many errors. This is most  likely due to model's inability to make generalization. As the training set size increases we can see that model is making less and less errors suggesting that model is improving in make generalization and fitting and predicting new data points. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "4cabe331-e5e7-41c4-9d92-ebd4538fdb8c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 0.4)"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x396 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.style.use('seaborn')\n",
    "plt.plot(train_sizes, train_scores_mean, label = 'Training error')\n",
    "plt.plot(train_sizes, validation_scores_mean, label = 'Validation error')\n",
    "plt.ylabel('MSE', fontsize = 14)\n",
    "plt.xlabel('Training set size', fontsize = 14)\n",
    "plt.title('Learning curves for SVM model', fontsize = 18, y = 1.03)\n",
    "plt.legend()\n",
    "plt.ylim(0,0.4)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e8b51f97-9ca3-48ff-b9e3-081bc84d13ab",
   "metadata": {},
   "source": [
    "The learning curve plot further supports out finding. Up to size 120, both training and validation errors decrease but there is a fairly big gap between the two lines suggesting that we are dealing with low bias and high variance. However when the training set size reaches 140, this gap considerably narrows down, indicating that our model has a low variance. Also after 140 point, there is not much change in either validation or training error meaning that adding more  data points to the training  wont improve the model's performance significantly. \n",
    "\n",
    "having a low bias and variance indicates that our model has optimum complexity "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "e5a4e4d1-b2ed-4bcc-93b0-229cbef89504",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['SVM_model.pkl']"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# writing out the model\n",
    "joblib.dump(cv.best_estimator_, 'SVM_model.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2109f427-86b2-49f2-81cf-81844e916511",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### **Multi Layered Perceptron (MLP)**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "39614a96-1c5e-4f3e-9b33-f76215613656",
   "metadata": {},
   "outputs": [],
   "source": [
    "import joblib\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore', category=FutureWarning)\n",
    "warnings.filterwarnings('ignore', category=DeprecationWarning)\n",
    "\n",
    "tr_features = pd.read_csv('train_features.csv')\n",
    "tr_labels = pd.read_csv('train_labels.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ed2be3db-1019-4cff-abfd-71cfd9640490",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Hyperparameter tuning: hidden_layer_sizes, Activation, learning_rate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "a257fd4a-1abe-48be-84df-45f4f1459b75",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BEST PARAMS: {'activation': 'logistic', 'hidden_layer_sizes': (100,), 'learning_rate': 'constant'}\n",
      "\n",
      "0.399 (+/-0.237) for {'activation': 'relu', 'hidden_layer_sizes': (10,), 'learning_rate': 'constant'}\n",
      "0.365 (+/-0.158) for {'activation': 'relu', 'hidden_layer_sizes': (10,), 'learning_rate': 'invscaling'}\n",
      "0.34 (+/-0.28) for {'activation': 'relu', 'hidden_layer_sizes': (10,), 'learning_rate': 'adaptive'}\n",
      "0.5 (+/-0.261) for {'activation': 'relu', 'hidden_layer_sizes': (50,), 'learning_rate': 'constant'}\n",
      "0.344 (+/-0.231) for {'activation': 'relu', 'hidden_layer_sizes': (50,), 'learning_rate': 'invscaling'}\n",
      "0.389 (+/-0.109) for {'activation': 'relu', 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive'}\n",
      "0.379 (+/-0.061) for {'activation': 'relu', 'hidden_layer_sizes': (100,), 'learning_rate': 'constant'}\n",
      "0.407 (+/-0.148) for {'activation': 'relu', 'hidden_layer_sizes': (100,), 'learning_rate': 'invscaling'}\n",
      "0.349 (+/-0.113) for {'activation': 'relu', 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive'}\n",
      "0.486 (+/-0.247) for {'activation': 'tanh', 'hidden_layer_sizes': (10,), 'learning_rate': 'constant'}\n",
      "0.427 (+/-0.022) for {'activation': 'tanh', 'hidden_layer_sizes': (10,), 'learning_rate': 'invscaling'}\n",
      "0.427 (+/-0.022) for {'activation': 'tanh', 'hidden_layer_sizes': (10,), 'learning_rate': 'adaptive'}\n",
      "0.543 (+/-0.285) for {'activation': 'tanh', 'hidden_layer_sizes': (50,), 'learning_rate': 'constant'}\n",
      "0.489 (+/-0.25) for {'activation': 'tanh', 'hidden_layer_sizes': (50,), 'learning_rate': 'invscaling'}\n",
      "0.486 (+/-0.3) for {'activation': 'tanh', 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive'}\n",
      "0.524 (+/-0.251) for {'activation': 'tanh', 'hidden_layer_sizes': (100,), 'learning_rate': 'constant'}\n",
      "0.592 (+/-0.28) for {'activation': 'tanh', 'hidden_layer_sizes': (100,), 'learning_rate': 'invscaling'}\n",
      "0.457 (+/-0.209) for {'activation': 'tanh', 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive'}\n",
      "0.427 (+/-0.022) for {'activation': 'logistic', 'hidden_layer_sizes': (10,), 'learning_rate': 'constant'}\n",
      "0.466 (+/-0.251) for {'activation': 'logistic', 'hidden_layer_sizes': (10,), 'learning_rate': 'invscaling'}\n",
      "0.413 (+/-0.05) for {'activation': 'logistic', 'hidden_layer_sizes': (10,), 'learning_rate': 'adaptive'}\n",
      "0.603 (+/-0.452) for {'activation': 'logistic', 'hidden_layer_sizes': (50,), 'learning_rate': 'constant'}\n",
      "0.665 (+/-0.438) for {'activation': 'logistic', 'hidden_layer_sizes': (50,), 'learning_rate': 'invscaling'}\n",
      "0.732 (+/-0.364) for {'activation': 'logistic', 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive'}\n",
      "0.878 (+/-0.44) for {'activation': 'logistic', 'hidden_layer_sizes': (100,), 'learning_rate': 'constant'}\n",
      "0.725 (+/-0.52) for {'activation': 'logistic', 'hidden_layer_sizes': (100,), 'learning_rate': 'invscaling'}\n",
      "0.649 (+/-0.554) for {'activation': 'logistic', 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive'}\n"
     ]
    }
   ],
   "source": [
    "mlp = MLPClassifier(max_iter = 2500)\n",
    "parameters = {\n",
    "    'hidden_layer_sizes': [(10,), (50,), (100,)],\n",
    "    'activation': ['relu', 'tanh', 'logistic'],\n",
    "    'learning_rate': ['constant', 'invscaling', 'adaptive']\n",
    "}\n",
    "\n",
    "cv = GridSearchCV(mlp, parameters, cv=5)\n",
    "cv.fit(tr_features, tr_labels.values.ravel())\n",
    "\n",
    "print_results(cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "684cf124-7fb1-483d-ab8d-7167ed5ae720",
   "metadata": {},
   "source": [
    "We can see that logistic activation generally results in higher performance. Together with hidden layer size of 100 and constant learning rate the MLP model has the performance of 87.8%. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "2f7c1958-e101-40e6-86f0-49e2c2d24fb4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MLPClassifier(activation='logistic', max_iter=2500)"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "05dbe850-a73a-40c8-86e1-79625d5c9998",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Learning Curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "8dc35abb-0ab7-41a6-a451-3799be1bb3ef",
   "metadata": {},
   "outputs": [],
   "source": [
    "#citation 3\n",
    "from sklearn.model_selection import learning_curve\n",
    "# the training data is 60% of the whole data --> 206\n",
    "# we are doing 5-fold cross validation with 80% training set and 20% validation set:\n",
    "# 206 training example : 80% training (164) 20% validation (42) in cross validation\n",
    "train_sizes = [5, 20, 40, 60, 80, 100, 120, 140, 164]\n",
    "train_sizes, train_scores, validation_scores = learning_curve(\n",
    "estimator = MLPClassifier(activation='logistic', hidden_layer_sizes=100, learning_rate='invscaling', max_iter=2500),\n",
    "X = tr_features,\n",
    "y = tr_labels.values.ravel(), train_sizes = train_sizes, cv = 5,\n",
    "scoring = 'neg_mean_squared_error')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "664a1581-76ea-4400-8e06-b9fc22cb34ff",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean training scores\n",
      "\n",
      " 5      0.240000\n",
      "20     0.480000\n",
      "40     0.725000\n",
      "60     0.336667\n",
      "80     1.315000\n",
      "100    0.438000\n",
      "120    0.648333\n",
      "140    0.091429\n",
      "164    0.712195\n",
      "dtype: float64\n",
      "\n",
      " --------------------\n",
      "\n",
      "Mean validation scores\n",
      "\n",
      " 5      0.713240\n",
      "20     0.750174\n",
      "40     0.824158\n",
      "60     0.341347\n",
      "80     1.318815\n",
      "100    0.390012\n",
      "120    0.658420\n",
      "140    0.097445\n",
      "164    0.694425\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "train_scores_mean = -train_scores.mean(axis = 1)\n",
    "validation_scores_mean = -validation_scores.mean(axis = 1)\n",
    "print('Mean training scores\\n\\n', pd.Series(train_scores_mean, index = train_sizes))\n",
    "print('\\n', '-' * 20) # separator\n",
    "print('\\nMean validation scores\\n\\n',pd.Series(validation_scores_mean, index = train_sizes))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5b8838f6-35d2-40a6-a1db-db615f4b12c3",
   "metadata": {},
   "source": [
    "We cannot see any clear pattern from the number except that error rates are oscillating quite a lot. This might be because our dataset is quite small and MLP is not suitable for small amount of data and mostly used for larger data. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "694a2a8d-ab77-42da-9f6a-8527670dc9f0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 2.0)"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x396 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.style.use('seaborn')\n",
    "plt.plot(train_sizes, train_scores_mean, label = 'Training error')\n",
    "plt.plot(train_sizes, validation_scores_mean, label = 'Validation error')\n",
    "plt.ylabel('MSE', fontsize = 14)\n",
    "plt.xlabel('Training set size', fontsize = 14)\n",
    "plt.title('Learning curves for MLP model', fontsize = 18, y = 1.03)\n",
    "plt.legend()\n",
    "plt.ylim(0,2)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1a63e7bf-251e-4915-8934-980d7599b5aa",
   "metadata": {},
   "source": [
    "the plot show that both training and validation error scores are following the same pattern. Neither of them is decreasing therefore the model has high bias. However both training and validation are following the same pattern and the gap between them is small. This suggests two things. First, the model has low variance and second adding extra data point will not help the model to improve and make better prediction meaning that the MLP is not suitable for our problem and we should consider other algorithms. \n",
    "\n",
    "Since MLP model has high bias and low variance we can conclude that this model is under fitting our data. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "id": "36e33bad-0a67-41eb-af3a-4840a84e077b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['MLP_model.pkl']"
      ]
     },
     "execution_count": 129,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "joblib.dump(cv.best_estimator_, 'MLP_model.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9a295f5b-c8e0-4cf9-a8c6-6d263117821b",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### **RF**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "41deea8c-d69b-456b-8bd2-4d1f06034d7f",
   "metadata": {},
   "outputs": [],
   "source": [
    "import joblib\n",
    "import pandas as pd\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore', category=FutureWarning)\n",
    "warnings.filterwarnings('ignore', category=DeprecationWarning)\n",
    "\n",
    "tr_features = pd.read_csv('train_features.csv')\n",
    "tr_labels = pd.read_csv('train_labels.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f8e9475d-07ba-47d2-adda-cf5733242e49",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Hyperparameter tuning: n_estimators, max_depth"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "e33eed60-861a-4e61-b418-b9c081591db8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BEST PARAMS: {'max_depth': 8, 'n_estimators': 250}\n",
      "\n",
      "0.952 (+/-0.043) for {'max_depth': 2, 'n_estimators': 5}\n",
      "0.966 (+/-0.023) for {'max_depth': 2, 'n_estimators': 50}\n",
      "0.971 (+/-0.019) for {'max_depth': 2, 'n_estimators': 250}\n",
      "0.976 (+/-0.03) for {'max_depth': 4, 'n_estimators': 5}\n",
      "0.976 (+/-0.03) for {'max_depth': 4, 'n_estimators': 50}\n",
      "0.976 (+/-0.03) for {'max_depth': 4, 'n_estimators': 250}\n",
      "0.981 (+/-0.036) for {'max_depth': 8, 'n_estimators': 5}\n",
      "0.981 (+/-0.036) for {'max_depth': 8, 'n_estimators': 50}\n",
      "0.986 (+/-0.038) for {'max_depth': 8, 'n_estimators': 250}\n",
      "0.971 (+/-0.019) for {'max_depth': 16, 'n_estimators': 5}\n",
      "0.981 (+/-0.036) for {'max_depth': 16, 'n_estimators': 50}\n",
      "0.976 (+/-0.03) for {'max_depth': 16, 'n_estimators': 250}\n",
      "0.976 (+/-0.043) for {'max_depth': 32, 'n_estimators': 5}\n",
      "0.981 (+/-0.036) for {'max_depth': 32, 'n_estimators': 50}\n",
      "0.981 (+/-0.036) for {'max_depth': 32, 'n_estimators': 250}\n",
      "0.947 (+/-0.056) for {'max_depth': None, 'n_estimators': 5}\n",
      "0.981 (+/-0.036) for {'max_depth': None, 'n_estimators': 50}\n",
      "0.981 (+/-0.036) for {'max_depth': None, 'n_estimators': 250}\n"
     ]
    }
   ],
   "source": [
    "rf = RandomForestClassifier()\n",
    "parameters = {\n",
    "    'n_estimators': [5, 50, 250],\n",
    "    'max_depth': [2, 4, 8, 16, 32, None]\n",
    "}\n",
    "\n",
    "cv = GridSearchCV(rf, parameters, cv=5)\n",
    "cv.fit(tr_features, tr_labels.values.ravel())\n",
    "\n",
    "print_results(cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ea21a8a6-c3dd-41b4-b4b6-0bcc4302c7d5",
   "metadata": {},
   "source": [
    "both n_estimator parameter and max_depth are affecting the performance of mode with n-estimators having a greater effect. When max_depth = 8 and there are 250 estimators, the model achieves its highest performance of 98.6%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "e4d810bd-a0ff-407b-ae73-ccd0b301977c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(max_depth=8, n_estimators=250)"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "777a0fa3-3e02-4433-93d6-105b8f8c7076",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Learning Curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "id": "d5249791-6f6a-4c43-97ee-91f18ca2af85",
   "metadata": {},
   "outputs": [],
   "source": [
    "#citation 3\n",
    "from sklearn.model_selection import learning_curve\n",
    "# the training data is 60% of the whole data --> 206\n",
    "# we are doing 5-fold cross validation with 80% training set and 20% validation set:\n",
    "# 206 training example : 80% training (164) 20% validation (42) in cross validation\n",
    "train_sizes = [5, 20, 40, 60, 80, 100, 120, 140, 164]\n",
    "train_sizes, train_scores, validation_scores = learning_curve(\n",
    "estimator = RandomForestClassifier(max_depth=8, n_estimators=250),\n",
    "X = tr_features,\n",
    "y = tr_labels.values.ravel(), train_sizes = train_sizes, cv = 5,\n",
    "scoring = 'neg_mean_squared_error')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "147ff137-5730-4f09-b511-a1fa35577f03",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean training scores\n",
      "\n",
      " 5     -0.0\n",
      "20    -0.0\n",
      "40    -0.0\n",
      "60    -0.0\n",
      "80    -0.0\n",
      "100   -0.0\n",
      "120   -0.0\n",
      "140   -0.0\n",
      "164   -0.0\n",
      "dtype: float64\n",
      "\n",
      " --------------------\n",
      "\n",
      "Mean validation scores\n",
      "\n",
      " 5      0.255981\n",
      "20     0.057840\n",
      "40     0.057840\n",
      "60     0.028920\n",
      "80     0.038676\n",
      "100    0.043554\n",
      "120    0.019280\n",
      "140    0.019280\n",
      "164    0.014402\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "train_scores_mean = -train_scores.mean(axis = 1)\n",
    "validation_scores_mean = -validation_scores.mean(axis = 1)\n",
    "print('Mean training scores\\n\\n', pd.Series(train_scores_mean, index = train_sizes))\n",
    "print('\\n', '-' * 20) # separator\n",
    "print('\\nMean validation scores\\n\\n',pd.Series(validation_scores_mean, index = train_sizes))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fa2da7b7-6c0f-4ec2-9632-a6d50704d48d",
   "metadata": {},
   "source": [
    "When in training, the model is able to fit and predict the data perfectly regardless of the data size. when tested on the validation set we can see that for the smaller size of training set, the models make many errors suggesting that the model is unable to make generalizations. As the size of the training size increases, the errors the model makes decreases indicating that the model is improving and is making more precise predictions. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "id": "c43c27b2-3541-43cf-ab93-fec1503703e6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 0.4)"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x396 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.style.use('seaborn')\n",
    "plt.plot(train_sizes, train_scores_mean, label = 'Training error')\n",
    "plt.plot(train_sizes, validation_scores_mean, label = 'Validation error')\n",
    "plt.ylabel('MSE', fontsize = 14)\n",
    "plt.xlabel('Training set size', fontsize = 14)\n",
    "plt.title('Learning curves for RF model', fontsize = 18, y = 1.03)\n",
    "plt.legend()\n",
    "plt.ylim(0,0.4)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "24857190-72aa-4863-84cb-c51d5a77432d",
   "metadata": {},
   "source": [
    "The plot also support our findings. Furthermore the low value of training error suggests that our model has low bias. However since the gap between the training and validation error is high, our model has a high variance compared to previous models. Since both lines remain different from each other and don't follow the same pattern, increasing the number of data point in our training might improve the model's  performance and it is one thing that we should consider.  \n",
    "\n",
    "the high variance and low bias suggest that RF model is over fitting our data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "2c93ade0-6010-42f5-90b7-1fa5aba15735",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['RF_model.pkl']"
      ]
     },
     "execution_count": 72,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "joblib.dump(cv.best_estimator_, 'RF_model.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5bd5d07d-92d5-4a18-9676-31fc284ff66b",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### **Boosting**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ba454758-3979-4dd8-84db-ee62ef13a61c",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Hyperparameter tuning: n-estimators, max_depth, & learning_rate Tuning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "id": "3cf9176d-f673-4611-a713-1ba5a2f0d62c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BEST PARAMS: {'learning_rate': 1, 'max_depth': 3, 'n_estimators': 5}\n",
      "\n",
      "0.427 (+/-0.022) for {'learning_rate': 0.01, 'max_depth': 1, 'n_estimators': 5}\n",
      "0.917 (+/-0.05) for {'learning_rate': 0.01, 'max_depth': 1, 'n_estimators': 50}\n",
      "0.966 (+/-0.023) for {'learning_rate': 0.01, 'max_depth': 1, 'n_estimators': 250}\n",
      "0.981 (+/-0.036) for {'learning_rate': 0.01, 'max_depth': 1, 'n_estimators': 500}\n",
      "0.427 (+/-0.022) for {'learning_rate': 0.01, 'max_depth': 3, 'n_estimators': 5}\n",
      "0.981 (+/-0.036) for {'learning_rate': 0.01, 'max_depth': 3, 'n_estimators': 50}\n",
      "0.981 (+/-0.036) for {'learning_rate': 0.01, 'max_depth': 3, 'n_estimators': 250}\n",
      "0.981 (+/-0.036) for {'learning_rate': 0.01, 'max_depth': 3, 'n_estimators': 500}\n",
      "0.427 (+/-0.022) for {'learning_rate': 0.01, 'max_depth': 5, 'n_estimators': 5}\n",
      "0.961 (+/-0.05) for {'learning_rate': 0.01, 'max_depth': 5, 'n_estimators': 50}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.01, 'max_depth': 5, 'n_estimators': 250}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.01, 'max_depth': 5, 'n_estimators': 500}\n",
      "0.427 (+/-0.022) for {'learning_rate': 0.01, 'max_depth': 7, 'n_estimators': 5}\n",
      "0.966 (+/-0.05) for {'learning_rate': 0.01, 'max_depth': 7, 'n_estimators': 50}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.01, 'max_depth': 7, 'n_estimators': 250}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.01, 'max_depth': 7, 'n_estimators': 500}\n",
      "0.427 (+/-0.022) for {'learning_rate': 0.01, 'max_depth': 9, 'n_estimators': 5}\n",
      "0.961 (+/-0.05) for {'learning_rate': 0.01, 'max_depth': 9, 'n_estimators': 50}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.01, 'max_depth': 9, 'n_estimators': 250}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.01, 'max_depth': 9, 'n_estimators': 500}\n",
      "0.917 (+/-0.05) for {'learning_rate': 0.1, 'max_depth': 1, 'n_estimators': 5}\n",
      "0.981 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 1, 'n_estimators': 50}\n",
      "0.976 (+/-0.052) for {'learning_rate': 0.1, 'max_depth': 1, 'n_estimators': 250}\n",
      "0.976 (+/-0.052) for {'learning_rate': 0.1, 'max_depth': 1, 'n_estimators': 500}\n",
      "0.981 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 3, 'n_estimators': 5}\n",
      "0.981 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 3, 'n_estimators': 50}\n",
      "0.981 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 3, 'n_estimators': 250}\n",
      "0.976 (+/-0.052) for {'learning_rate': 0.1, 'max_depth': 3, 'n_estimators': 500}\n",
      "0.966 (+/-0.05) for {'learning_rate': 0.1, 'max_depth': 5, 'n_estimators': 5}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 5, 'n_estimators': 50}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 5, 'n_estimators': 250}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 5, 'n_estimators': 500}\n",
      "0.961 (+/-0.05) for {'learning_rate': 0.1, 'max_depth': 7, 'n_estimators': 5}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 7, 'n_estimators': 50}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 7, 'n_estimators': 250}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 7, 'n_estimators': 500}\n",
      "0.961 (+/-0.05) for {'learning_rate': 0.1, 'max_depth': 9, 'n_estimators': 5}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 9, 'n_estimators': 50}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 9, 'n_estimators': 250}\n",
      "0.971 (+/-0.036) for {'learning_rate': 0.1, 'max_depth': 9, 'n_estimators': 500}\n",
      "0.976 (+/-0.03) for {'learning_rate': 1, 'max_depth': 1, 'n_estimators': 5}\n",
      "0.981 (+/-0.036) for {'learning_rate': 1, 'max_depth': 1, 'n_estimators': 50}\n",
      "0.981 (+/-0.036) for {'learning_rate': 1, 'max_depth': 1, 'n_estimators': 250}\n",
      "0.981 (+/-0.036) for {'learning_rate': 1, 'max_depth': 1, 'n_estimators': 500}\n",
      "0.985 (+/-0.024) for {'learning_rate': 1, 'max_depth': 3, 'n_estimators': 5}\n",
      "0.981 (+/-0.036) for {'learning_rate': 1, 'max_depth': 3, 'n_estimators': 50}\n",
      "0.981 (+/-0.036) for {'learning_rate': 1, 'max_depth': 3, 'n_estimators': 250}\n",
      "0.981 (+/-0.036) for {'learning_rate': 1, 'max_depth': 3, 'n_estimators': 500}\n",
      "0.961 (+/-0.066) for {'learning_rate': 1, 'max_depth': 5, 'n_estimators': 5}\n",
      "0.961 (+/-0.066) for {'learning_rate': 1, 'max_depth': 5, 'n_estimators': 50}\n",
      "0.966 (+/-0.049) for {'learning_rate': 1, 'max_depth': 5, 'n_estimators': 250}\n",
      "0.966 (+/-0.049) for {'learning_rate': 1, 'max_depth': 5, 'n_estimators': 500}\n",
      "0.976 (+/-0.031) for {'learning_rate': 1, 'max_depth': 7, 'n_estimators': 5}\n",
      "0.961 (+/-0.066) for {'learning_rate': 1, 'max_depth': 7, 'n_estimators': 50}\n",
      "0.966 (+/-0.049) for {'learning_rate': 1, 'max_depth': 7, 'n_estimators': 250}\n",
      "0.966 (+/-0.049) for {'learning_rate': 1, 'max_depth': 7, 'n_estimators': 500}\n",
      "0.961 (+/-0.066) for {'learning_rate': 1, 'max_depth': 9, 'n_estimators': 5}\n",
      "0.966 (+/-0.049) for {'learning_rate': 1, 'max_depth': 9, 'n_estimators': 50}\n",
      "0.966 (+/-0.049) for {'learning_rate': 1, 'max_depth': 9, 'n_estimators': 250}\n",
      "0.966 (+/-0.049) for {'learning_rate': 1, 'max_depth': 9, 'n_estimators': 500}\n",
      "0.485 (+/-0.324) for {'learning_rate': 10, 'max_depth': 1, 'n_estimators': 5}\n",
      "0.485 (+/-0.324) for {'learning_rate': 10, 'max_depth': 1, 'n_estimators': 50}\n",
      "0.485 (+/-0.324) for {'learning_rate': 10, 'max_depth': 1, 'n_estimators': 250}\n",
      "0.485 (+/-0.324) for {'learning_rate': 10, 'max_depth': 1, 'n_estimators': 500}\n",
      "0.947 (+/-0.046) for {'learning_rate': 10, 'max_depth': 3, 'n_estimators': 5}\n",
      "0.956 (+/-0.047) for {'learning_rate': 10, 'max_depth': 3, 'n_estimators': 50}\n",
      "0.952 (+/-0.043) for {'learning_rate': 10, 'max_depth': 3, 'n_estimators': 250}\n",
      "0.961 (+/-0.038) for {'learning_rate': 10, 'max_depth': 3, 'n_estimators': 500}\n",
      "0.966 (+/-0.023) for {'learning_rate': 10, 'max_depth': 5, 'n_estimators': 5}\n",
      "0.971 (+/-0.036) for {'learning_rate': 10, 'max_depth': 5, 'n_estimators': 50}\n",
      "0.971 (+/-0.048) for {'learning_rate': 10, 'max_depth': 5, 'n_estimators': 250}\n",
      "0.971 (+/-0.036) for {'learning_rate': 10, 'max_depth': 5, 'n_estimators': 500}\n",
      "0.971 (+/-0.019) for {'learning_rate': 10, 'max_depth': 7, 'n_estimators': 5}\n",
      "0.971 (+/-0.036) for {'learning_rate': 10, 'max_depth': 7, 'n_estimators': 50}\n",
      "0.971 (+/-0.048) for {'learning_rate': 10, 'max_depth': 7, 'n_estimators': 250}\n",
      "0.971 (+/-0.036) for {'learning_rate': 10, 'max_depth': 7, 'n_estimators': 500}\n",
      "0.966 (+/-0.023) for {'learning_rate': 10, 'max_depth': 9, 'n_estimators': 5}\n",
      "0.971 (+/-0.036) for {'learning_rate': 10, 'max_depth': 9, 'n_estimators': 50}\n",
      "0.971 (+/-0.048) for {'learning_rate': 10, 'max_depth': 9, 'n_estimators': 250}\n",
      "0.966 (+/-0.049) for {'learning_rate': 10, 'max_depth': 9, 'n_estimators': 500}\n",
      "0.451 (+/-0.212) for {'learning_rate': 100, 'max_depth': 1, 'n_estimators': 5}\n",
      "0.451 (+/-0.212) for {'learning_rate': 100, 'max_depth': 1, 'n_estimators': 50}\n",
      "0.451 (+/-0.212) for {'learning_rate': 100, 'max_depth': 1, 'n_estimators': 250}\n",
      "0.451 (+/-0.212) for {'learning_rate': 100, 'max_depth': 1, 'n_estimators': 500}\n",
      "0.956 (+/-0.057) for {'learning_rate': 100, 'max_depth': 3, 'n_estimators': 5}\n",
      "0.956 (+/-0.057) for {'learning_rate': 100, 'max_depth': 3, 'n_estimators': 50}\n",
      "0.956 (+/-0.057) for {'learning_rate': 100, 'max_depth': 3, 'n_estimators': 250}\n",
      "0.956 (+/-0.057) for {'learning_rate': 100, 'max_depth': 3, 'n_estimators': 500}\n",
      "0.956 (+/-0.036) for {'learning_rate': 100, 'max_depth': 5, 'n_estimators': 5}\n",
      "0.961 (+/-0.024) for {'learning_rate': 100, 'max_depth': 5, 'n_estimators': 50}\n",
      "0.951 (+/-0.031) for {'learning_rate': 100, 'max_depth': 5, 'n_estimators': 250}\n",
      "0.961 (+/-0.05) for {'learning_rate': 100, 'max_depth': 5, 'n_estimators': 500}\n",
      "0.971 (+/-0.036) for {'learning_rate': 100, 'max_depth': 7, 'n_estimators': 5}\n",
      "0.942 (+/-0.024) for {'learning_rate': 100, 'max_depth': 7, 'n_estimators': 50}\n",
      "0.961 (+/-0.05) for {'learning_rate': 100, 'max_depth': 7, 'n_estimators': 250}\n",
      "0.961 (+/-0.039) for {'learning_rate': 100, 'max_depth': 7, 'n_estimators': 500}\n",
      "0.966 (+/-0.023) for {'learning_rate': 100, 'max_depth': 9, 'n_estimators': 5}\n",
      "0.966 (+/-0.023) for {'learning_rate': 100, 'max_depth': 9, 'n_estimators': 50}\n",
      "0.971 (+/-0.036) for {'learning_rate': 100, 'max_depth': 9, 'n_estimators': 250}\n",
      "0.961 (+/-0.05) for {'learning_rate': 100, 'max_depth': 9, 'n_estimators': 500}\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "\n",
    "gb = GradientBoostingClassifier()\n",
    "parameters = {\n",
    "    'n_estimators': [5, 50, 250, 500],\n",
    "    'max_depth': [1, 3, 5, 7, 9],\n",
    "    'learning_rate': [0.01, 0.1, 1, 10, 100]\n",
    "}\n",
    "\n",
    "cv = GridSearchCV(gb, parameters, cv=5)\n",
    "cv.fit(tr_features, tr_labels.values.ravel())\n",
    "\n",
    "print_results(cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "693b6c07-a4e4-4ad1-861d-1914b61204fc",
   "metadata": {},
   "source": [
    "overall the model has the best performance (98.5%) with learning rate of 1, max_depth of 3 and 5 estimators. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "id": "d776798f-e759-4aa1-9788-adde709869c4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "GradientBoostingClassifier(learning_rate=1, n_estimators=5)"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "20774057-2571-423c-accf-74d3e12975f8",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Learning Curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "id": "5214cc32-7209-4b49-8707-078600d121f2",
   "metadata": {},
   "outputs": [],
   "source": [
    "#citation 3\n",
    "from sklearn.model_selection import learning_curve\n",
    "# the training data is 60% of the whole data --> 206\n",
    "# we are doing 5-fold cross validation with 80% training set and 20% validation set:\n",
    "# 206 training example : 80% training (164) 20% validation (42) in cross validation\n",
    "train_sizes = [5, 20, 40, 60, 80, 100, 120, 140, 164]\n",
    "train_sizes, train_scores, validation_scores = learning_curve(\n",
    "estimator = GradientBoostingClassifier(learning_rate=1, max_depth = 3, n_estimators=5),\n",
    "X = tr_features,\n",
    "y = tr_labels.values.ravel(), train_sizes = train_sizes, cv = 5,\n",
    "scoring = 'neg_mean_squared_error')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "id": "d243b728-40b4-4df2-8056-0314a27d2f46",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean training scores\n",
      "\n",
      " 5     -0.0\n",
      "20    -0.0\n",
      "40    -0.0\n",
      "60    -0.0\n",
      "80    -0.0\n",
      "100   -0.0\n",
      "120   -0.0\n",
      "140   -0.0\n",
      "164   -0.0\n",
      "dtype: float64\n",
      "\n",
      " --------------------\n",
      "\n",
      "Mean validation scores\n",
      "\n",
      " 5      0.503833\n",
      "20     0.169570\n",
      "40     0.082230\n",
      "60     0.076887\n",
      "80     0.038792\n",
      "100    0.038792\n",
      "120    0.057840\n",
      "140    0.067480\n",
      "164    0.024274\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "train_scores_mean = -train_scores.mean(axis = 1)\n",
    "validation_scores_mean = -validation_scores.mean(axis = 1)\n",
    "print('Mean training scores\\n\\n', pd.Series(train_scores_mean, index = train_sizes))\n",
    "print('\\n', '-' * 20) # separator\n",
    "print('\\nMean validation scores\\n\\n',pd.Series(validation_scores_mean, index = train_sizes))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "36e89ced-6574-438a-b5ec-4651467c1a18",
   "metadata": {},
   "source": [
    "0 training error suggest the model is able to fit and predict while being trained regardless of training set size. While being tested on the validation set, the model makes many errors for smaller sizes since it is unable of making generalization. As the size increase we can see that the model make less errors suggesting that it is getting better at making predictions. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "id": "fcae6b1c-dd20-4dcd-9310-4ee65436cc21",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 0.6)"
      ]
     },
     "execution_count": 77,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x396 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.style.use('seaborn')\n",
    "plt.plot(train_sizes, train_scores_mean, label = 'Training error')\n",
    "plt.plot(train_sizes, validation_scores_mean, label = 'Validation error')\n",
    "plt.ylabel('MSE', fontsize = 14)\n",
    "plt.xlabel('Training set size', fontsize = 14)\n",
    "plt.title('Learning curves for GB model', fontsize = 18, y = 1.03)\n",
    "plt.legend()\n",
    "plt.ylim(0,0.6)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5550b761-df9c-485a-8cc7-84eed5699f3b",
   "metadata": {},
   "source": [
    " The model has the lowest bias and variance when the training set size is set to 100. overall the model has a low bias but since the gap between training and validation is higher than other models, this model has higher variance. This means that the models is over fitting. \n",
    "\n",
    "Finally, since the training and validation set lines continue to differ and do not converge or become constant, we can conclude that adding more data points in training will most likely improve the model significantly. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "id": "5aa607f6-a88b-4821-8006-569b36ea3ed5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['GB_model.pkl']"
      ]
     },
     "execution_count": 78,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "joblib.dump(cv.best_estimator_, 'GB_model.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "85b2cfd1-1b1f-4f78-acfd-80545a2d4c4f",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### **Model comparison**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f5131feb-7f17-42f8-8dbb-cb4b1e171f81",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Read in Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "id": "ae671f51-7e08-4cd7-a863-0339bf9d13a3",
   "metadata": {},
   "outputs": [],
   "source": [
    "import joblib\n",
    "import pandas as pd\n",
    "from sklearn.metrics import accuracy_score, precision_score, recall_score\n",
    "from time import time\n",
    "\n",
    "val_features = pd.read_csv('val_features.csv')\n",
    "val_labels = pd.read_csv('val_labels.csv')\n",
    "\n",
    "te_features = pd.read_csv('test_features.csv')\n",
    "te_labels = pd.read_csv('test_labels.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "644f69a1-ab2f-4eee-9b8e-a71728c1c4ae",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Read in Models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "id": "b9b1dc2b-4f45-49fc-9722-fb225dab896c",
   "metadata": {},
   "outputs": [],
   "source": [
    "models = {}\n",
    "\n",
    "for mdl in ['LR', 'SVM', 'MLP', 'RF', 'GB']:\n",
    "    models[mdl] = joblib.load('{}_model.pkl'.format(mdl))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "id": "f6813aa3-4fef-450d-8c5b-d4a63c592457",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'LR': LogisticRegression(C=100, max_iter=3000),\n",
       " 'SVM': SVC(C=0.1, kernel='linear'),\n",
       " 'MLP': MLPClassifier(activation='logistic', max_iter=2500),\n",
       " 'RF': RandomForestClassifier(max_depth=8, n_estimators=250),\n",
       " 'GB': GradientBoostingClassifier(learning_rate=1, n_estimators=5)}"
      ]
     },
     "execution_count": 81,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "models"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2dec30e2-1682-465a-8475-635ae084f2bf",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Evaluate models on the validation set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "id": "99902119-7dda-493f-81a8-04f12f8de4c0",
   "metadata": {},
   "outputs": [],
   "source": [
    "def evaluate_model(name, model, features, labels):\n",
    "    start = time()\n",
    "    pred = model.predict(features)\n",
    "    end = time()\n",
    "    latency = round((end - start)*1000, 1)\n",
    "    accuracy = round(accuracy_score(labels, pred), 3)\n",
    "    precision = round(precision_score(labels, pred, average='macro'), 3)\n",
    "    recall = round(recall_score(labels, pred, average='macro'), 3)\n",
    "    print('{} --> Accuracy= {} | Precision= {} | Recall= {} | Latency= {}ms'.format(name, accuracy, precision, recall, latency ))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "id": "2a74f166-4e12-4a81-ac0b-6f90122ea4f7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR --> Accuracy= 0.986 | Precision= 0.987 | Recall= 0.99 | Latency= 1.0ms\n",
      "SVM --> Accuracy= 0.986 | Precision= 0.99 | Recall= 0.972 | Latency= 0.8ms\n",
      "MLP --> Accuracy= 0.681 | Precision= 0.454 | Recall= 0.56 | Latency= 1.0ms\n",
      "RF --> Accuracy= 0.986 | Precision= 0.974 | Recall= 0.99 | Latency= 21.0ms\n",
      "GB --> Accuracy= 0.986 | Precision= 0.974 | Recall= 0.99 | Latency= 1.0ms\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\elmir\\anaconda3\\lib\\site-packages\\sklearn\\metrics\\_classification.py:1248: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "for name, mdl in models.items():\n",
    "    evaluate_model(name, mdl, val_features, val_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9f90cbbe-d093-4dc7-a936-ee1e4a59a5dc",
   "metadata": {},
   "source": [
    "From the result above we can see that MLP with the accuracy of 46.4% is worst algorithm for the purpose of our project. From our discussion on the Learning curve of the algorithm, this was expected. We also discussed how MLP is underfitting while RF and GB are overfitting our data Leaving only LR and SVM. \n",
    "\n",
    "The other 4 algorithm have the same accuracy of 98.6%. Therefore we have to consider the precision/ recall trade off. \n",
    "\n",
    "In our project false negative is when model labels an example negative for belonging to one of the species while the example actually belongs to that species and false positive is when model predicts an example's species while it does not belong to that species. In this case having a false positive is worse than false negative (it is worse for model to say a penguin belongs to a certain species and it doesn't than to say a penguin doesn't belong to a species when in fact it does) therefore precision has higher priority than recall. \n",
    "\n",
    "that being said, the SVM with accuracy of 98.6% , precision of 99%, recall of 97.2% and latency of 0.8ms is the best model to predict the species of the penguins. Coming second is LR model with the same accuracy, precision of 98.7% and recall 99%. Additionally since our dataset is quite small, SVM would be more suitable to use than other models. In the next step, we are going to run our model on testing set to investigate our hypothesis's correctness. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "768cdc3e-df1e-43fa-baf5-ffba2fa589ab",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Evaluate models on test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "id": "13d5aafc-bffd-4b63-a659-0f8f415482f7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR --> Accuracy= 0.986 | Precision= 0.987 | Recall= 0.99 | Latency= 2.0ms\n",
      "SVM --> Accuracy= 0.986 | Precision= 0.99 | Recall= 0.987 | Latency= 1.5ms\n",
      "MLP --> Accuracy= 0.623 | Precision= 0.415 | Recall= 0.515 | Latency= 1.0ms\n",
      "RF --> Accuracy= 0.971 | Precision= 0.964 | Recall= 0.976 | Latency= 21.0ms\n",
      "GB --> Accuracy= 0.971 | Precision= 0.964 | Recall= 0.976 | Latency= 1.0ms\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\elmir\\anaconda3\\lib\\site-packages\\sklearn\\metrics\\_classification.py:1248: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "for name, mdl in models.items():\n",
    "    evaluate_model(name, mdl, te_features, te_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e010b261-cc1e-43cc-a533-9979a78595db",
   "metadata": {},
   "source": [
    "After running the model on the test set, we can see that the accuracy score dropped for RF and GB models making LR and SVM models with highest accuracy. Both models maintained the same precision but their latency almost doubled. furthermore, the Recall of the SVM model increase to 98.7% while the recall of LR remained same. Since there are no significant improvement in LR model, We can confidently say that when splitting data 60-20-20, SVM model is the best model with accuracy of 98.6%, precision of 99%, recall of 98.7% and latency of 1.5 ms. \n",
    "\n",
    "in the next section we are going to explore the models performance with split ratio of 80:10:10 \n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d0412486-d255-43b0-8fc4-65a2a4b20aee",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "## **Part IV: Splitting data 80-10-10**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "eef303f2-bb32-4b20-974f-bdfa009e7b58",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### Reading cleaned data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "id": "f41c5bea-011e-4fb9-9b99-1ddfdc64bfdc",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>island</th>\n",
       "      <th>culmen_length_mm</th>\n",
       "      <th>culmen_depth_mm</th>\n",
       "      <th>flipper_length_mm</th>\n",
       "      <th>body_mass_g</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>39.10000</td>\n",
       "      <td>18.70000</td>\n",
       "      <td>181.000000</td>\n",
       "      <td>3750.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>39.50000</td>\n",
       "      <td>17.40000</td>\n",
       "      <td>186.000000</td>\n",
       "      <td>3800.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>40.30000</td>\n",
       "      <td>18.00000</td>\n",
       "      <td>195.000000</td>\n",
       "      <td>3250.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>43.92193</td>\n",
       "      <td>17.15117</td>\n",
       "      <td>200.915205</td>\n",
       "      <td>4201.754386</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>36.70000</td>\n",
       "      <td>19.30000</td>\n",
       "      <td>193.000000</td>\n",
       "      <td>3450.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   species  island  culmen_length_mm  culmen_depth_mm  flipper_length_mm  \\\n",
       "0        0       1          39.10000         18.70000         181.000000   \n",
       "1        0       1          39.50000         17.40000         186.000000   \n",
       "2        0       1          40.30000         18.00000         195.000000   \n",
       "3        0       1          43.92193         17.15117         200.915205   \n",
       "4        0       1          36.70000         19.30000         193.000000   \n",
       "\n",
       "   body_mass_g  \n",
       "0  3750.000000  \n",
       "1  3800.000000  \n",
       "2  3250.000000  \n",
       "3  4201.754386  \n",
       "4  3450.000000  "
      ]
     },
     "execution_count": 85,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "data = pd.read_csv('penguins_cleaned.csv')\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "93b62f74-4de3-4c91-9c8b-eef0f5b282dc",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### Splitting data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "id": "d8989e31-a5f2-4baa-bf8f-089705adb498",
   "metadata": {},
   "outputs": [],
   "source": [
    "features = data.drop('species', axis=1)\n",
    "labels = data['species']\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(features, labels, test_size=0.2, random_state=42)\n",
    "X_test, X_val, y_test, y_val = train_test_split(X_test, y_test, test_size=0.5, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "id": "38a71c18-97cd-473e-802f-b6a613e79430",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8\n",
      "0.1\n",
      "0.1\n"
     ]
    }
   ],
   "source": [
    "for dataset in [y_train, y_val, y_test]:\n",
    "    print(round(len(dataset) / len(labels), 2))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "26988018-2eaf-46a2-9727-f271c5e48b84",
   "metadata": {},
   "source": [
    "Saving the three datasets:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "id": "07087b5d-71db-4649-9b09-4c9215f41d50",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train.to_csv('train_features2.csv', index=False)\n",
    "X_val.to_csv('val_features2.csv', index=False)\n",
    "X_test.to_csv('test_features2.csv', index=False)\n",
    "\n",
    "y_train.to_csv('train_labels2.csv', index=False)\n",
    "y_val.to_csv('val_labels2.csv', index=False)\n",
    "y_test.to_csv('test_labels2.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fa3c1268-4f1c-47db-b2e4-9c3337d42dd5",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### **Logistic regression algorithm:**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "id": "288d433e-a59d-4690-bfea-01ef834a336e",
   "metadata": {},
   "outputs": [],
   "source": [
    "import joblib\n",
    "import pandas as pd\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.model_selection import GridSearchCV # used for tuning algorithms and hyperparameters\n",
    "from sklearn.svm import SVC\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore', category=FutureWarning)\n",
    "warnings.filterwarnings('ignore', category=DeprecationWarning)\n",
    "\n",
    "tr_features = pd.read_csv('train_features2.csv')\n",
    "tr_labels = pd.read_csv('train_labels2.csv')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fd6e94cc-bd9d-4c95-a14a-4df3bff1e053",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Hyperparameter tuning: C"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "id": "21ab541c-2056-4713-a97b-716efc668d41",
   "metadata": {},
   "outputs": [],
   "source": [
    "def print_results(results):\n",
    "    print('BEST PARAMS: {}\\n'.format(results.best_params_))\n",
    "\n",
    "    means = results.cv_results_['mean_test_score']\n",
    "    stds = results.cv_results_['std_test_score']\n",
    "    for mean, std, params in zip(means, stds, results.cv_results_['params']):\n",
    "        print('{} (+/-{}) for {}'.format(round(mean, 3), round(std * 2, 3), params))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "id": "d43b89f9-16d5-44d6-ad9d-df89214a95bc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BEST PARAMS: {'C': 1}\n",
      "\n",
      "0.905 (+/-0.048) for {'C': 0.001}\n",
      "0.967 (+/-0.036) for {'C': 0.01}\n",
      "0.978 (+/-0.036) for {'C': 0.1}\n",
      "0.982 (+/-0.04) for {'C': 1}\n",
      "0.978 (+/-0.053) for {'C': 10}\n",
      "0.982 (+/-0.04) for {'C': 100}\n",
      "0.982 (+/-0.04) for {'C': 1000}\n"
     ]
    }
   ],
   "source": [
    "lr = LogisticRegression(max_iter=4500)\n",
    "parameters = {\n",
    "    'C': [0.001, 0.01, 0.1, 1, 10, 100, 1000]\n",
    "}\n",
    "\n",
    "cv = GridSearchCV(lr, parameters, cv=5)\n",
    "cv.fit(tr_features, tr_labels.values.ravel())\n",
    "\n",
    "print_results(cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f8f86eca-7b4a-4588-b4e4-8d0187b348db",
   "metadata": {},
   "source": [
    "With C = 1 the model being trained on 80% of the dataset has the performance of 98.2%  which is less than performance rate of when model was being trained on 60% of the dataset proving our hypothesis from the previous section on how increasing the data point in the training set will not improve the model's performance. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "id": "bcdc15bd-c3b0-4817-b6df-b0a20e94f177",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=1, max_iter=4500)"
      ]
     },
     "execution_count": 98,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "45e0832b-cb7b-4763-b43a-cd201d7d3efb",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Learning Curve:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "id": "866a535d-550c-4b54-8105-2af853584c34",
   "metadata": {},
   "outputs": [],
   "source": [
    "#citation 3\n",
    "from sklearn.model_selection import learning_curve\n",
    "# the training data is 80% of the whole data --> 275\n",
    "# we are doing 5-fold cross validation with 80% training set and 20% validation set:\n",
    "# 275 training example : 80% training (220) 20% validation (55) in cross validation\n",
    "train_sizes = [5, 50, 100, 150, 200, 220]\n",
    "train_sizes, train_scores, validation_scores = learning_curve(\n",
    "estimator = LogisticRegression(C=1, max_iter=4500),\n",
    "X = tr_features,\n",
    "y = tr_labels.values.ravel(), train_sizes = train_sizes, cv = 5,\n",
    "scoring = 'neg_mean_squared_error')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "id": "942f8b5a-fb1e-4b30-bca9-c753904fc11a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean training scores\n",
      "\n",
      " 5     -0.000000\n",
      "50    -0.000000\n",
      "100   -0.000000\n",
      "150   -0.000000\n",
      "200   -0.000000\n",
      "220    0.000909\n",
      "dtype: float64\n",
      "\n",
      " --------------------\n",
      "\n",
      "Mean validation scores\n",
      "\n",
      " 5      0.843636\n",
      "50     0.036364\n",
      "100    0.047273\n",
      "150    0.021818\n",
      "200    0.047273\n",
      "220    0.040000\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "train_scores_mean = -train_scores.mean(axis = 1)\n",
    "validation_scores_mean = -validation_scores.mean(axis = 1)\n",
    "print('Mean training scores\\n\\n', pd.Series(train_scores_mean, index = train_sizes))\n",
    "print('\\n', '-' * 20) # separator\n",
    "print('\\nMean validation scores\\n\\n',pd.Series(validation_scores_mean, index = train_sizes))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f4d73282-b57e-41d4-a470-bf98939b9d04",
   "metadata": {},
   "source": [
    "As when we split 60:20:20, We can see the model is able to fit and predict perfectly when in training regardless of the training set size. Except when the training set size is at its highest (220) where model makes some error. This is probably the reason why the overall model performance  is lower than when the split ration was 60:20:20. \n",
    "\n",
    "The model follows the same pattern as last time when being tested in validation mode, making more errors because of its inability to make correct prediction due to small training size and improving as the training size increases. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "id": "5d4d7f69-8438-4b6d-858c-9cff48656dcf",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 1.0)"
      ]
     },
     "execution_count": 101,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x396 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.style.use('seaborn')\n",
    "plt.plot(train_sizes, train_scores_mean, label = 'Training error')\n",
    "plt.plot(train_sizes, validation_scores_mean, label = 'Validation error')\n",
    "plt.ylabel('MSE', fontsize = 14)\n",
    "plt.xlabel('Training set size', fontsize = 14)\n",
    "plt.title('Learning curves for a LR model', fontsize = 18, y = 1.03)\n",
    "plt.legend()\n",
    "plt.ylim(0,1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5f271143-b978-4c68-b873-b0eca6112610",
   "metadata": {},
   "source": [
    "The learning curve further support our findings. Additionally we can see that the model still has a low bias but it has a higher variance than previous split ratio. So even though the model has optimum complexity it slightly shifts towards over fitting in compared to previous LR model. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "id": "69a52c36-fcc0-4baa-b60d-5499a53b7610",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['LR_model2.pkl']"
      ]
     },
     "execution_count": 103,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "joblib.dump(cv.best_estimator_, 'LR_model2.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d9162561-46c2-44e6-904b-15beb0f07816",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### **Support Vector Machine (SVM)**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "id": "0c789c37-d6d4-4201-b5e1-c18da73eed54",
   "metadata": {},
   "outputs": [],
   "source": [
    "import joblib\n",
    "import pandas as pd\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.model_selection import GridSearchCV # used for tuning algorithms and hyperparameters\n",
    "from sklearn.svm import SVC\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore', category=FutureWarning)\n",
    "warnings.filterwarnings('ignore', category=DeprecationWarning)\n",
    "\n",
    "tr_features = pd.read_csv('train_features2.csv')\n",
    "tr_labels = pd.read_csv('train_labels2.csv')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5cab8e30-1572-4217-aead-1c2f4836e65c",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### C & Kernel tuning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "id": "06354172-9a2f-497b-a246-174905d8d6dd",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BEST PARAMS: {'C': 0.1, 'kernel': 'linear'}\n",
      "\n",
      "0.989 (+/-0.018) for {'C': 0.1, 'kernel': 'linear'}\n",
      "0.745 (+/-0.065) for {'C': 0.1, 'kernel': 'rbf'}\n",
      "0.982 (+/-0.023) for {'C': 1, 'kernel': 'linear'}\n",
      "0.742 (+/-0.084) for {'C': 1, 'kernel': 'rbf'}\n",
      "0.985 (+/-0.027) for {'C': 10, 'kernel': 'linear'}\n",
      "0.745 (+/-0.076) for {'C': 10, 'kernel': 'rbf'}\n"
     ]
    }
   ],
   "source": [
    "svc = SVC()\n",
    "parameters = { # used to search for the hyperparameters\n",
    "    'kernel': ['linear', 'rbf'],\n",
    "    'C': [0.1, 1, 10]\n",
    "}\n",
    "\n",
    "cv = GridSearchCV(svc, parameters, cv=5) # cv = 5 --> 5 fold validation\n",
    "cv.fit(tr_features, tr_labels.values.ravel())\n",
    "\n",
    "print_results(cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4b5ab12b-9887-4068-aa1b-bdc9d53c3fe2",
   "metadata": {},
   "source": [
    "Based on the above values, we can see that kernel has a great impact on the performance of the model and linear kernel is doing relatively better than rbf suggesting that our data is linearly separable. We can also see that the performance of the three values of C are fairly similar with minor difference but overall the linear kernel with C = 0.1 has the best performance with 98.9% performance rate which is slightly lower than when model was trained on 60:20:20 ratio. \n",
    "\n",
    "This decrease in supports our hypothesis that further increasing the data point in the training will not improve the SVM model's performance. As a matter of fact, it decrease the performance. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "id": "d8323d5c-9ec4-43c1-b281-9e369dba4066",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SVC(C=0.1, kernel='linear')"
      ]
     },
     "execution_count": 107,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "60114896-b809-4109-983c-a1bd7a5d69bb",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Learning Curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "id": "1c9babdf-342d-43f0-8e2c-94bdbd757c13",
   "metadata": {},
   "outputs": [],
   "source": [
    "#citation 3\n",
    "from sklearn.model_selection import learning_curve\n",
    "# the training data is 80% of the whole data --> 275\n",
    "# we are doing 5-fold cross validation with 80% training set and 20% validation set:\n",
    "# 275 training example : 80% training (220) 20% validation (55) in cross validation\n",
    "train_sizes = [5, 50, 100, 150, 200, 220]\n",
    "train_sizes, train_scores, validation_scores = learning_curve(\n",
    "estimator = SVC(C=0.1, kernel='linear'),\n",
    "X = tr_features,\n",
    "y = tr_labels.values.ravel(), train_sizes = train_sizes, cv = 5,\n",
    "scoring = 'neg_mean_squared_error')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "id": "85670430-c477-412a-9555-3be5490a2958",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean training scores\n",
      "\n",
      " 5     -0.000000\n",
      "50     0.016000\n",
      "100    0.040000\n",
      "150    0.026667\n",
      "200    0.023000\n",
      "220    0.020909\n",
      "dtype: float64\n",
      "\n",
      " --------------------\n",
      "\n",
      "Mean validation scores\n",
      "\n",
      " 5      0.843636\n",
      "50     0.094545\n",
      "100    0.036364\n",
      "150    0.021818\n",
      "200    0.021818\n",
      "220    0.021818\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "train_scores_mean = -train_scores.mean(axis = 1)\n",
    "validation_scores_mean = -validation_scores.mean(axis = 1)\n",
    "print('Mean training scores\\n\\n', pd.Series(train_scores_mean, index = train_sizes))\n",
    "print('\\n', '-' * 20) # separator\n",
    "print('\\nMean validation scores\\n\\n',pd.Series(validation_scores_mean, index = train_sizes))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ce375bec-fe38-4ec4-a677-64361d70ba61",
   "metadata": {},
   "source": [
    "As in the last split ratio, the model is able to make prefect fitting and prediction when the training set has the size of 5. However it has high errors when being tested on validation set indicating that model is unable to make generalization when the training set size is too small. As the training set size increase the model makes more error while training but makes better predictions when being tested on the validation set. \n",
    "\n",
    "one thing of note is when size  = 150, the training error is 2.67% while the validation error is 2.18%. This pattern continues for the larger training set, further supporting our hypothesis that the increasing the training data for SVM model not only does not improve the model but in fact impacts its performance negatively. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "id": "0d9793e8-8a71-45d3-af55-6a84ff0c4c98",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 1.0)"
      ]
     },
     "execution_count": 111,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x396 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.style.use('seaborn')\n",
    "plt.plot(train_sizes, train_scores_mean, label = 'Training error')\n",
    "plt.plot(train_sizes, validation_scores_mean, label = 'Validation error')\n",
    "plt.ylabel('MSE', fontsize = 14)\n",
    "plt.xlabel('Training set size', fontsize = 14)\n",
    "plt.title('Learning curves for SVM model', fontsize = 18, y = 1.03)\n",
    "plt.legend()\n",
    "plt.ylim(0,1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "25177c56-0d02-4108-a8f1-5e510b407522",
   "metadata": {},
   "source": [
    "In the plot we can see that training error is quite low suggesting that our model has low bias. Since the gap between training error and validation error is quite low (and might be negative is some places) therefore, our model also has low variance.\n",
    "\n",
    "This suggest that our model has optimum complexity with a slight shift towards underfitting compared to previous SVM model. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "id": "14a7b703-061c-47e7-b2f1-ca45706f1686",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['SVM_model2.pkl']"
      ]
     },
     "execution_count": 112,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# writing out the model\n",
    "joblib.dump(cv.best_estimator_, 'SVM_model2.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5b6a3b2d-25b0-4c63-8fbb-a317013695f0",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### **MLP**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "id": "63b1e401-cb79-45f4-b152-a408e75fcb41",
   "metadata": {},
   "outputs": [],
   "source": [
    "import joblib\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore', category=FutureWarning)\n",
    "warnings.filterwarnings('ignore', category=DeprecationWarning)\n",
    "\n",
    "tr_features = pd.read_csv('train_features2.csv')\n",
    "tr_labels = pd.read_csv('train_labels2.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c9f88be9-b337-46c9-abdf-466078b0ec12",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Hyperparameter tuning: hidden_layer_sizes, activation, learning_rate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "id": "2dafd356-5300-4e79-a332-74fae321891a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BEST PARAMS: {'activation': 'logistic', 'hidden_layer_sizes': (100,), 'learning_rate': 'constant'}\n",
      "\n",
      "0.364 (+/-0.188) for {'activation': 'relu', 'hidden_layer_sizes': (10,), 'learning_rate': 'constant'}\n",
      "0.4 (+/-0.26) for {'activation': 'relu', 'hidden_layer_sizes': (10,), 'learning_rate': 'invscaling'}\n",
      "0.262 (+/-0.232) for {'activation': 'relu', 'hidden_layer_sizes': (10,), 'learning_rate': 'adaptive'}\n",
      "0.375 (+/-0.224) for {'activation': 'relu', 'hidden_layer_sizes': (50,), 'learning_rate': 'constant'}\n",
      "0.422 (+/-0.334) for {'activation': 'relu', 'hidden_layer_sizes': (50,), 'learning_rate': 'invscaling'}\n",
      "0.462 (+/-0.235) for {'activation': 'relu', 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive'}\n",
      "0.404 (+/-0.302) for {'activation': 'relu', 'hidden_layer_sizes': (100,), 'learning_rate': 'constant'}\n",
      "0.451 (+/-0.363) for {'activation': 'relu', 'hidden_layer_sizes': (100,), 'learning_rate': 'invscaling'}\n",
      "0.371 (+/-0.305) for {'activation': 'relu', 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive'}\n",
      "0.484 (+/-0.189) for {'activation': 'tanh', 'hidden_layer_sizes': (10,), 'learning_rate': 'constant'}\n",
      "0.436 (+/-0.0) for {'activation': 'tanh', 'hidden_layer_sizes': (10,), 'learning_rate': 'invscaling'}\n",
      "0.436 (+/-0.0) for {'activation': 'tanh', 'hidden_layer_sizes': (10,), 'learning_rate': 'adaptive'}\n",
      "0.491 (+/-0.218) for {'activation': 'tanh', 'hidden_layer_sizes': (50,), 'learning_rate': 'constant'}\n",
      "0.436 (+/-0.0) for {'activation': 'tanh', 'hidden_layer_sizes': (50,), 'learning_rate': 'invscaling'}\n",
      "0.487 (+/-0.241) for {'activation': 'tanh', 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive'}\n",
      "0.509 (+/-0.285) for {'activation': 'tanh', 'hidden_layer_sizes': (100,), 'learning_rate': 'constant'}\n",
      "0.655 (+/-0.1) for {'activation': 'tanh', 'hidden_layer_sizes': (100,), 'learning_rate': 'invscaling'}\n",
      "0.625 (+/-0.224) for {'activation': 'tanh', 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive'}\n",
      "0.487 (+/-0.204) for {'activation': 'logistic', 'hidden_layer_sizes': (10,), 'learning_rate': 'constant'}\n",
      "0.436 (+/-0.0) for {'activation': 'logistic', 'hidden_layer_sizes': (10,), 'learning_rate': 'invscaling'}\n",
      "0.502 (+/-0.262) for {'activation': 'logistic', 'hidden_layer_sizes': (10,), 'learning_rate': 'adaptive'}\n",
      "0.491 (+/-0.276) for {'activation': 'logistic', 'hidden_layer_sizes': (50,), 'learning_rate': 'constant'}\n",
      "0.607 (+/-0.214) for {'activation': 'logistic', 'hidden_layer_sizes': (50,), 'learning_rate': 'invscaling'}\n",
      "0.436 (+/-0.0) for {'activation': 'logistic', 'hidden_layer_sizes': (50,), 'learning_rate': 'adaptive'}\n",
      "0.716 (+/-0.049) for {'activation': 'logistic', 'hidden_layer_sizes': (100,), 'learning_rate': 'constant'}\n",
      "0.655 (+/-0.233) for {'activation': 'logistic', 'hidden_layer_sizes': (100,), 'learning_rate': 'invscaling'}\n",
      "0.596 (+/-0.281) for {'activation': 'logistic', 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive'}\n"
     ]
    }
   ],
   "source": [
    "mlp = MLPClassifier(max_iter = 2500)\n",
    "parameters = {\n",
    "    'hidden_layer_sizes': [(10,), (50,), (100,)],\n",
    "    'activation': ['relu', 'tanh', 'logistic'],\n",
    "    'learning_rate': ['constant', 'invscaling', 'adaptive']\n",
    "}\n",
    "\n",
    "cv = GridSearchCV(mlp, parameters, cv=5)\n",
    "cv.fit(tr_features, tr_labels.values.ravel())\n",
    "\n",
    "print_results(cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "af59ee14-b393-4725-80d9-098f3b377e77",
   "metadata": {},
   "source": [
    "Same as previous split ratio, logistic activation generally results in higher performance. Together with hidden layer size of 100 and constant learning rate, for the split ratio of 80:10:10,  the MLP model has the performance of 71.6% which is quite lower than model's performance in the previous section. As we discussed in the previous section, the MLP model is not very suitable for our problem and increasing the training data point will not make any difference at this time. \n",
    "\n",
    "the only thing that might improve the model's performance is to gather more data (e.g., instead of 344 we need like 1000 data point)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "id": "1120ff3a-d2d2-4487-8501-96ac3f46e19d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MLPClassifier(activation='logistic', max_iter=2500)"
      ]
     },
     "execution_count": 124,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "82e59e7a-d6bd-4e40-8551-b4352635e445",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Learning Curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "id": "e9902b65-4228-4398-b136-ca4867e74f25",
   "metadata": {},
   "outputs": [],
   "source": [
    "#citation 3\n",
    "from sklearn.model_selection import learning_curve\n",
    "# the training data is 80% of the whole data --> 275\n",
    "# we are doing 5-fold cross validation with 80% training set and 20% validation set:\n",
    "# 275 training example : 80% training (220) 20% validation (55) in cross validation\n",
    "train_sizes = [5, 50, 100, 150, 200, 220]\n",
    "train_sizes, train_scores, validation_scores = learning_curve(\n",
    "estimator = MLPClassifier(activation='logistic', max_iter=2500),\n",
    "X = tr_features,\n",
    "y = tr_labels.values.ravel(), train_sizes = train_sizes, cv = 5,\n",
    "scoring = 'neg_mean_squared_error')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "id": "20cadf85-4319-47e4-845c-859a16bc309a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean training scores\n",
      "\n",
      " 5      0.480000\n",
      "50     0.712000\n",
      "100    1.408000\n",
      "150    0.634667\n",
      "200    0.973000\n",
      "220    0.563636\n",
      "dtype: float64\n",
      "\n",
      " --------------------\n",
      "\n",
      "Mean validation scores\n",
      "\n",
      " 5      0.930909\n",
      "50     0.745455\n",
      "100    1.418182\n",
      "150    0.654545\n",
      "200    0.974545\n",
      "220    0.669091\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "train_scores_mean = -train_scores.mean(axis = 1)\n",
    "validation_scores_mean = -validation_scores.mean(axis = 1)\n",
    "print('Mean training scores\\n\\n', pd.Series(train_scores_mean, index = train_sizes))\n",
    "print('\\n', '-' * 20) # separator\n",
    "print('\\nMean validation scores\\n\\n',pd.Series(validation_scores_mean, index = train_sizes))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "id": "6784d620-d008-4f1d-a1ad-a1435bd58850",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 2.0)"
      ]
     },
     "execution_count": 127,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfkAAAF4CAYAAABXWoCZAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAABsdUlEQVR4nO3deXhMZ/vA8e9kjyRkkQSJRBKCbLLZ963U0lJLqNKWt1WlC91o+7baKrovtIqWotpautGidkppLVkkdiFkkX2b7JM5vz/yml9TRJDMJJP7c12uy5z1njtn5p5zznOeR6UoioIQQgghjI6JoQMQQgghRO2QIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyotbMnj2btm3bkpiYaOhQbsu1uEXtOHXqFA888ACBgYH069eP2niKNzExkbZt29K2bVvmzZt30+V27NihW+7acXpt3dmzZ99yP9fW/ec/f39/unXrxjPPPMOFCxdq7D3VtIkTJ9KvXz+9rScMw8zQAQhR10RERNC1a1dDh2G0XnnlFS5evMisWbNo2rQpKpWqVve3a9cuXn311RvO27lz511v39vbmyeeeEL3ury8nKSkJNasWcPBgwf56aefaNmy5V3vR4g7IUVeiH8JCQkhJCTE0GEYrbNnz9K3b18effTRWt+Xu7s7iYmJnDp1ivbt21eap9Fo2Lt3L46OjmRlZd3xPpo2bcr9999/3fSgoCAef/xxli9fzptvvnnH2xfibsjleiGEXpWVlWFjY6OXffXt2xdTU1N27dp13bwjR46Qk5ND3759a2XfvXv3xtbWlsjIyFrZvhDVIUVe1Annz59n+vTphIeH06FDB8aNG8cff/xx3XLbtm3joYceIiwsjICAAPr168e7775LaWmpbpmJEycyZcoUPvroI0JCQujatStnzpzRTd+/f7/unnCfPn1YtGgRWq1Wt/6/78nPnj2bwYMHExMTw0MPPUSHDh3o1q0b8+bNo7i4uFJ88fHxTJs2jfDwcDp37sy8efNYv359tdomqNVq5s+fT58+fejQoQPDhw9nw4YNuvmLFi264Xb+PX3RokUEBgayY8cOunfvTkhICMuWLaNt27asXLnyuv3Onj2bkJAQioqKAMjNzeWtt96iZ8+eBAQEcO+997Jq1arr7p1/9913DB8+nA4dOtC5c2emT5/OuXPnbvr+fvzxR11ef/rpJ9q2bcuPP/4IQFFRER988AH9+vXT/V3ff/99XUz/XP/333+nX79+dOjQgUWLFlWZ0yZNmhAWFsbu3buvm7djxw4CAwNp3rx5ldu4GyYmJpSXl990/uzZsxk2bBjHjh0jIiKCoKAg+vfvz08//URZWRkffPAB3bt3p1OnTjz77LNkZ2dXWv/MmTM8+eSThIeHExQUxNixY294C+LPP/9k3LhxBAcHM2DAALZs2XLDeKr7ORT1hxR5YXBnzpwhIiKC8+fPM3XqVGbOnIlGo+Hxxx+v9GW0YcMGnnnmGezs7Hj++ed58cUXcXNz46uvvmLZsmWVtnn8+HF+++03XnjhBUaOHEnr1q2BikvFzz77LJ07d+bVV1+lZcuWLF68mO+++67KGLOyspgyZQre3t688sorhIaGsmbNGj799FPdMsnJyTz44INERkYyefJkpkyZwo4dO/jggw9umYPS0lImTJjAN998Q58+fZgzZw7u7u68+uqrrF69+nbSCVRcin711Vd5+OGHmTJlCgMGDMDT05OtW7det9+dO3cyYMAArK2tKSws5KGHHuKXX35h5MiRvPzyy7Rp04b58+dXuuS8adMm5s6di5+fH6+88gqPPvoox48fZ+LEieTn598wpo4dO/Luu+8CEB4ezrvvvkvHjh0pLS3l0UcfZfny5XTp0oWXX36ZTp06sXz5ciZPnkxZWVml7cyZM4fhw4fz1FNPVavtxIABA4iLiyMlJUU3TVEUdu7cycCBA6ud09sVGxtLXl4efn5+VS6Xnp7OE088QVhYGC+99BJmZma8/PLLTJ06lcOHD/Pkk08ybNgwtm7dqssfQExMDBEREcTExPDoo48ya9YsysrKmD59OmvXrtUt9+eff/LYY4+Rn5/Ps88+y5AhQ3jllVeIi4urFEd1P4einlGEqCUvvfSS4uvrq1y5cqXK5R566CFlwIABSkFBgW5aWVmZ8uCDDyrdunVTSkpKFEVRlMGDBysRERGKVquttFyvXr2UYcOGVdqer6+vcvjw4ev24+vrq+zatUs3rbi4WOnYsaMSERFxXdz/fr169epK27v33nuVHj166F7PmTNH8fPzU86fP6+bdvXqVSU4OPiWeVi7dq3i6+urbNq0STdNq9UqDz74oNK9e3dFo9Eon3766Q238+/p115/+umnlZb75JNPFF9fXyUpKUk3befOnYqvr6+yb98+3br+/v7K6dOnK637wQcfKL6+vsqpU6cURVGU//znP8rQoUMrLbN3715lyJAhytGjR2/6PhVFUXx9fZWXXnpJ9/rbb79VfH19lZUrV1Zabvny5Yqvr6+ydu1aRVEU5Ycffrhu3Zu5cuWKLgeJiYmKr6+v8s033+jmR0dHK76+vsqFCxeuy9+1dauzH19fXyUiIkLJzMzU/bt8+bKybds2pX///oqfn59y4sSJm65/7dhas2aNbtrevXsVX19fpW/fvrpjX1EUZdy4cZWOtzFjxijBwcFKSkqKblpxcbEycuRIJSgoSMnMzFQURVFGjhyp9O7dW8nPz9ctd+jQId0+rqnu5/Chhx6qtJ6o2+RMXhhUdnY2f//9N71796a4uJisrCyysrLIy8tj4MCBZGRkcOLECaDi7HHZsmWVWmNnZmbSuHFjCgsLK23XysqKjh07Xrc/a2tr+vTpo3ttaWmJl5cXGRkZt4z13nvvrfS6Xbt2ZGZmAhVnhrt27aJnz574+PjolnF1deW+++675bavNQAbNmyYbppKpeLdd99l7dq1mJjc/ke1R48elV4PHz4cqLjlcc2WLVtwcnKiW7duAGzfvh1fX1+cnZ11f4usrCwGDBgAwJ49ewBo1qwZ8fHxLF68WHeboHfv3vz222+EhYXdVpy7d+/G1taWCRMmVJo+adIkbG1tr7uf/u/3dStubm60b9++0nZ27NiBj48P3t7et7WtG4mMjKRr1666fwMGDODpp59Gq9XyySefEBAQcMtt/POKQqtWrQDo2bMnFhYWuunu7u6kp6cDkJGRQXR0NPfffz/NmjXTLWNpacmUKVMoLi7mzz//JDMzk7i4OIYOHYqtra1uuS5dulS6JXU7n0NRv0jremFQV65cAWDNmjWsWbPmhstcu8xqbm7OkSNH+PXXX4mPj+fy5cu6Iuvm5lZpHXt7+xsWxhtNt7CwqHRP/mYcHR2vW+/a/dacnBxycnJ0X9D/VJ1CkpSUhIeHx3WPk/37fd0OJyenSq+9vLzw9/dn27ZtTJ48meLiYnbv3s2oUaMwM6v4Krh8+TLFxcU3vQx+7W8xffp0oqKiWLRoEYsWLaJ169b069ePMWPG4OHhcVtxJiYm0rJlS8zNzStNt7CwoGXLliQlJVX5vqqjf//+LF26FLVaja2tLTt27GDQoEG3vZ0b+fcz9RYWFjg7O+Pp6VntbfzzPZmaml437dp05X/tIq7lxMvL67ptXfuRmZycrFvuRn8Tb29vYmJigNv7HIr6RYq8MKhrRXLChAm6s8V/u3Y//YMPPmDZsmX4+fkRHBzM/fffT0hICG+99dZ1X0DXvij/7U7OiKuzrkajAah05nWNpaXlLbddXl5+x8+L36xh143ive+++1iwYAFJSUmcOHGCwsLCSlcPysvLCQsLY8aMGTfcpouLC1BxJv/LL7/w119/sWvXLv744w+WLVvGypUrWbFiBZ06dap2/EoVneFotdrriv+d/A0HDBjA4sWL2b9/P23btuXixYvcc889t72dG2nSpInuSsiduvYj65+qOh5ulTOo+FF8bRslJSU3XQ5u73Mo6hcp8sKgrp2pmpqaXvdFef78eRITE7G2tiYpKYlly5Zx//33V2p8BFTrUnttc3JyolGjRly6dOm6eQkJCbdcv0WLFpw5c+a66fv27WPLli288MILuuL2zycJ4Pbe/5AhQ3jnnXfYtWsXx44do2XLlgQHB+vmu7m5UVBQcN3fIjc3l0OHDunOTq/Feu0SNcCxY8d4+OGHWbNmzW0VeTc3N6KioigrK6tU0EtLS0lMTCQ8PLza27qZ9u3b4+bmxq5du7hy5Qpubm74+/vf9XYN5drnJj4+/rp5Fy9eBCp+iLm5uaFSqW54XP7zKY3qfg5F/SP35IVBubi4EBAQwE8//URqaqpuellZGS+//DJPP/00Go2G3Nxc4PqziX379nHp0iXdmbShmJiY0K9fP/bv36+79AkVxfHXX3+95fq9evUiIyODHTt2VJq+atUq9u7di4ODA87OzgCcPn1aN1+tVrNv375qx+ni4kKXLl3YsWMH+/fv192nv6Zfv36cPn2avXv3Vpq+ZMkSnnnmGd0jcs888wwvvvhipasIfn5+mJub3/aZdr9+/VCr1ZVahAN8++23FBQUVGpDcTf69+/PgQMH2LFjx03PVusLZ2dnAgIC2LRpE1evXtVNLy0tZeXKlVhYWNC9e3ccHR3p2LEjmzZtqvRjMDIyslLr+up+DkX9I2fyotZ99NFHN+z85N5776Vr1666R71GjRrF+PHjsbe357fffiM6OprnnnsOBwcHbGxsaNGiBV988QUlJSU0a9aMmJgYfvrpJywtLSkoKDDAO6vsmWeeYd++fURERDBx4kQsLCz4/vvvycvLA6q+/Dpu3Dh++OEHZs6cyYQJE/Dy8mLv3r0cPHiQ+fPnY2pqyoABA5g3bx5vvvkmSUlJWFhYsH79eho1anRbcQ4fPpw5c+YAVLpUDzB16lS2b9/OjBkzGDduHG3atOHYsWP88ssv9OrVi169egEwZcoUXn31VR555BEGDx6Moij88ssvlJSU8OCDD95WPGPGjOGnn35i4cKFnD17loCAAGJjY/nxxx/p0KEDY8aMua3t3cyAAQNYvXo1OTk5vPTSS7dcPjIyktdee+266dbW1rr8GdK1z83o0aMZP348NjY2bNq0ibi4OF599VUaN24MwEsvvcSECRMYO3YsEyZMoKioiK+//hoHB4cbbq+qz6Gof6TIi1p3szNZb29vunbtSkhICN999x2LFi1i5cqVaDQavLy8WLhwISNHjgQq7nUvW7aMhQsXsnr1ahRFwcPDg5dffhmNRsPbb79NbGxstVoy1xYPDw+++eYb3nnnHZYuXYqlpSUjRozA1NSUr7766ob366+xsrJizZo1fPzxx/z222/k5+fj4+PDxx9/rGvV7+joyPLly/nggw/49NNPcXBwYOzYsXh7ezNz5sxqx3nPPfcwd+5cWrduXelJAKhomLhu3To+/fRTtm3bxrp162jRogVPPvkkjz/+uO4sfcyYMZibm7N69Wo+/PBDtFotAQEBLF++nM6dO99W3iwsLPj666/57LPP2Lp1K5s2baJZs2ZMnTqVadOmXXdP/k6Fh4djb2+PqalptZ4AuHTp0g0vc9vZ2dWJIn/tc/Ppp5+yYsUKtFot7dq147PPPqt0pSIgIIA1a9bwwQcfsHjxYho3bsyMGTOIjY3l+PHj122vqs+hqH9USlUtOIQQ1ZaZmYmjo+N1Z+xvvfUW3333HdHR0TVWsIQQojrknrwQNeSZZ55h6NChlVotFxUVsWfPHtq1aycFXgihd3K5Xogacv/99/Pqq6/y+OOP079/f0pKSnQNo9544w1DhyeEaIDkcr0QNWjTpk2sXr2a+Ph4TExMCAgI4Mknn7ytR8qEEKKmSJEXQgghjJTckxdCCCGMlBR5IYQQwkhJkRdCCCGMlBR5IYQQwkhJkRdCCCGMlBR5IYQQwkhJkRdCCCGMlF56vLs2XGFSUhKlpaVMmzaN/v376+bv3r2bzz77DDMzM0aNGsXYsWPRarXMnTuXM2fOYGFhwbx583RjWQshhBDi1vRS5Ddt2oS9vT3vvfce2dnZjBw5Ulfky8rKWLBgARs3bsTa2prx48fTt29fIiMjKS0tZd26dURFRbFw4UKWLFmij3CFEEIIo6CXIj948GAGDRqke21qaqr7/4ULF/Dw8KBJkyYAhIWFcfToUaKioujZsycAwcHBxMbG6iNUIYQQwmjopcjb2NgAoFarefrpp3n22Wd189RqNXZ2dpWWVavVqNVqbG1tddNNTU3RaDSYmVUdskZTjpmZaZXLCCGEEA2B3kahS0lJYfr06Tz44IMMHz5cN93W1paCggLd64KCAuzs7K6brtVqb1ngAbKzC6sdk7OzHenp+dVeXtQMybthSN4NR3JvGA0l787Odjedp5fW9RkZGUyePJkXXniB0aNHV5rn4+NDQkICOTk5lJaWcvToUUJCQggNDWX//v0AREVF4evrq49QhRBCCKOhlzP5L774gry8PD7//HM+//xzAMaMGUNRURERERHMnj2bKVOmoCgKo0aNwtXVlYEDB3Lw4EHGjRuHoijMnz9fH6EKIYQQRsPohpq9nUszDeVSTl0jeTcMybvhSO4No6Hk3eCX64UQQgihf1LkhRBCCCMlRV4IIYQwUlLkhRBCCCOlt+fkhRBCiOpYtOgjzpw5RVZWJsXFxbRo4Ya9vQPz5r1zy3XXrPmasLBw/PwCbjj/k08+ICJiAs2aNavpsOskaV3fAFpe1jWSd8OQvBuO5P7ObNmymYSES0yb9tQdrd9Q8l5V63o5kxdCCHFT63ef58jptBrdZsd2Lozt1/q213v77bnk5uaSl5fLO+98yJIli0hLSyU3N5cuXbrx2GPTePvtufTvfw9ZWZkcP/4XeXlqkpISmTDhYYYMGc6MGY/zwgsvs3Pn76SkJJOdnU1qagpPPTWLzp27cvDgH3z11RfY2NhiZ9cYH5/WTJkyVReDWq1m4cI3yc3NBeDZZ1/Ax6c1o0YNw9OzFZ6eXqjV+bo43333Y1at+oqYmCgABg4czNix4yu9l3ff/ZjGjRvXSG7/TYq8EEKIeiMsLJyIiAmkpCTj7x/I7Nn/paSkhAceGMJjj02rtKxarebddz/mypXLvPTSTIYMGV5pvrm5BR988ClHjhzmu+/WEh7eiY8/fp+lS1fg6OjEG2+8et3+V69eQVhYJ0aOHM2VK5eZP/8Nliz5irS0VFas+IYmTex5++25ujgPHvyDlJRkli37mvLycqZNm0JYWMdK76U2SZEXQghxU2P7tb6js+7a4uHhCUDjxo05dSqO48ePYmNjQ2lp2XXLtmvXDgAXF1dKS0uvm+/r2/Z/85tRWlpCTk42NjY2ODo6AdChQzCZmZmV1omPP8/x40fZtWs7APn5FbcDmjSxp0kT++viTEi4SIcOwahUKszMzPD3D+TSpfhKy9QmaV0vhBCi3lCpKsrWli2/Ymtrx+uvz2PcuIcoKSnm303MVCrVLbZV+bWDgyOFhQVkZ2cDEBd3/RDnnp6tGDv2QRYvXsZbby3knnsGA2BiUrmcXovT09NLd6leo9EQGxuDu7tHpWVqk5zJCyGEqHfCwjoyd+7LxMREYWVlhbt7SzIy0u9qmyYmJsyc+SIvvPAMNja2KIoWd/eWlZaZNGkyCxe+xaZNP1JYWMDkyY9Xuc3u3XsSGXmMqVMfpaysjH79BtC2bbu7ivN2SOv6BtDysq6RvBuG5N1wJPeGcSd5X7NmJRERE7CwsODNN/9Lx46duffeYbUUYc2Q1vVCCCFENTRq1IipUx/BysqKZs1a0L//PYYO6a5IkRdCCCH+Z9SoCEaNijB0GDVGGt4JIYQQRkqKvBBCCGGkpMgLIYQQRkqKvBBCCGGkpMgLIYSoU6ZPf4xjx45Umvbxx++zefPPN1x+9OjhlJSUsGbN15w8WbkDm5KSEkaPHn7D9a755Zcf0Wg0nDt3hpUrl99V7HWNFHkhhBB1yn33jWTbtt90r8vKyjh48A8GDBhU5XoTJz5y0yFmq7JmzUrKy8tp06Ytjz762G2vX5fJI3RCCCFu6sfzvxKZdqJGtxniEsgDrW/ewUyfPv1ZtuxziouLsbKy4o8/9tGpU2fy8/N4/fWXKS0tIS8vl0ceeYxevfro1rs2Al1QUDBvvvkqxcWFuLg0182PjDymO1MvLi7m1VffICYmkqysTObOfZkxY8bzyy8/8MYbC9i+fSvr13+Hubk5LVt68OKLr7B9+1YOHTpISUlxpZHt/mn37p2sW7cWExMTgoKCmTbtKb76aimxsTEUFRUxe/Z/ee212TRu3ISuXbvTsWNnPvroPUxNTbGwsODFF19FUbS89NJM3TITJjx8x7mWM3khhBB1iqWlJT179mb//j0AbNmyifvue4CEhEuMGzeBjz/+nJkzX+THH9ffcP2tWzfj5eXD2rVruf/+UbrpFy/G89prb/Hpp1/Qo0cv9uzZybBhI3B0dGLu3Pm65XJzc/jqq6V8+ukSliz5CltbW3755QcACgoqRrZbuPBDvvnm60r7zcvLZcWKpXzyScV6GRlpHDlyGKjow/6LL1ZgaWlJVlYmH330GRMmPMw777zNrFkvsnjxMkaOHM3ixR8CVFrmbsiZvBBCiJt6oPWwKs+6a8vw4SP57LNPCA0NJz8/n7Zt2xEff4FVq77it99+AVRoNJobrnvxYjydO3cFwN8/ADOzilLn7OzMxx+/h7V1I9LT0wgM7HDD9ZOTk/Dy8qZRIxsAOnQI5ciRw/j5BdC6tS9w45HtEhOvkJOTzfPPPw1AYWEhSUlJQOUR55o3b4G5uTkAGRnptGnTVrefL75YfN0yd0PO5IUQQtQ5Pj6tKSoqYP367xg69D4AvvzyCwYPHsp///sWoaHhN13Xw6MVsbEVtxjOnj2t+zHwzjvzePnl13nllbk0beqsW16lMqk0gl3z5m5cunSRoqIiAKKijtOy5bWR424+sl3z5m64uLjy8cefs3jxMkaPjsDfv6KNgInJ/6/3z9HnmjZ15vz5czfYT82UZzmTF0IIUScNHXofn332KT/88CsAffv255NP3mfNmpW4uLiSk5Nzw/UeeGAMCxa8wfjx42nRoqXujHjQoCE8/vgj2NnZ4eDgpBu1rkOHYJ5//mndiHL29vZMnjyVp5+eikplgrt7S554YoZuDPmbcXBwICJiAjNmPE55eTnNm7egX7+BVa7z0kuv8NFH76IoCqampsye/d/bSdEtySh0MjKU3kneDUPybjiSe8NoKHmvahQ6uVwvhBBCGCkp8kIIIYSRkiIvhBBCGCm9NryLjo7m/fffZ82aNbpp6enpzJo1S/f61KlTPPfcc4wfP54RI0ZgZ1dxr8Hd3Z0FCxboM1whhBCiXtNbkV++fDmbNm3C2tq60nRnZ2dd0Y+MjOSjjz5i7NixlJSUAFT6QSCEEEKI6tPb5XoPDw8WLVp00/mKovDWW28xd+5cTE1NOX36NEVFRUyePJlJkyYRFRWlr1CFEEIIo6C3M/lBgwaRmJh40/m7d++mTZs2eHt7A2BlZcWUKVMYM2YMly5d4rHHHmPbtm26notuxsGhEWZmptWOq6pHD0TtkbwbhuTdcCT3htHQ815nOsPZtGkTkyZN0r328vLC09MTlUqFl5cX9vb2pKen07x58yq2AtnZhdXeZ0N5hrKukbwbhuTdcCT3htFQ8l4vnpOPi4sjNDRU93rjxo0sXLgQgNTUVNRqNc7OzjdbXQghhBD/YrAiv3nzZtatWwdAVlYWNjY2lfoEHj16NPn5+YwfP56ZM2cyf/78W16qF0IIIcT/k25tG8ClnLpG8m4YknfDkdwbRkPJe724XC+EEEKImiVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIIIyVFXgghhDBSUuSFEEIII6XXIh8dHc3EiROvm75y5UqGDh3KxIkTmThxIvHx8Wi1Wl577TUiIiKYOHEiCQkJ+gxVCCGEqPfM9LWj5cuXs2nTJqytra+bFxcXxzvvvENAQIBu2vbt2yktLWXdunVERUWxcOFClixZoq9whRBCiHpPb2fyHh4eLFq06Ibz4uLiWLZsGePHj2fp0qUAHDt2jJ49ewIQHBxMbGysvkIVQgghjILezuQHDRpEYmLiDecNHTqUBx98EFtbW2bMmMGePXtQq9XY2trqljE1NUWj0WBmpreQhRBCiHrN4BVTURQefvhh7OzsAOjduzcnT57E1taWgoIC3XJarbZaBd7BoRFmZqbV3r+zs93tBy3umuTdMCTvhiO5N4yGnneDF3m1Ws2wYcPYsmULjRo14q+//mLUqFEUFxezZ88ehgwZQlRUFL6+vtXaXnZ2YbX37exsR3p6/p2GLu6Q5N0wJO+GI7k3jIaS96p+yBisyG/evJnCwkIiIiKYOXMmkyZNwsLCgq5du9K7d2+0Wi0HDx5k3LhxKIrC/PnzDRWqEEIIUS+pFEVRDB1ETbqdX20N5VdeXSN5NwzJu+FI7g2joeS9qjN56QxHCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeiAaiXFtu6BCEEHomRV6IBmDbpV088tNzRKadMHQoQgg9kiIvhJHbfeUPNsf/TommhFUnv+dKfpKhQxJC6IkUeSGM2KHkI/xwbjMqjRWlCe0pKy9jaczX5JYYfy9gQggp8kIYreNpMaw9vRFVuQVFp8JoXOhLWaIv2SW5LD+xirLyMkOHKISoZVLkhTBCJzPP8HXcd6A1peh0GF192vDZi/1wKPZDk9Gci3mX+e7MjxjZ0BVCiH+RIi+EkTmfc5FlJ1aj1SoUnwkhsJk3j9zbDrtGFjwzqgOmSR1QCprw19Vj7Lqy39DhCiFqkRR5IYzIlfwklkSvRKPVUHwuGC87L6bdH4CZacVHvUVTGx4fHkTJ2RAos+Ln81uIzThl4KiFELVFirwQRiK1II3FUV9SXF5MyflAXE1b8cyYDlhamFZaLrh1Ux7o5kfxmRDQmrAi7luS1VcNFLUQojZJkRfCCGQWZfNp1HLUZQWUXvSjcZkXs8YGY2ttfsPlh3TxpKNnG0riAygpL2FpzNeoywr0HLUQorZJkReinssrzWdR1DJySnLRXGmLVb43s8Z2wKmJ1U3XUalUPDqkPW7mbShL8iGjOIsvT6yRXvGEMDJS5IWoxwrLClkc9SXpRZlor/qgSvfhmdEdcHO2veW6luamPPVAENbZ7SnPduVcTjwbzm3SQ9RCCH2RIi9EPVWsKeHz6JUkqVMgw5OyK2144v4AWrs3qfY2nJpYMX1kEOUXg6CoMX8kHWJ/4p+1GLUQQp+kyAtRD5WVl7H8xGou5iVgmutOUXw7Hr63HcFtmt72tnxb2vPQgIqGeKpySzac3cTprHO1ELUQQt+kyAtRz5Rry1kZ9y2ns89hXtAc9Vk/RvX2oWdQizveZu9gN/oGtKb4TDBaBb6K/Ya0wowajFoIYQhS5IWoR7SKlrWnNxKdEYdlsQt5JwMYGObJkC6ed73tcf3b4OvoRelFPwo1RXwR8zVFmqIaiFoIYShS5IWoJxRFYeO5zfx19RhWZU7kxAbRuX0LIvq3RqVS3fX2zUxNmDYiAIey1pSltCK1MI0Vsd+iVbQ1EL0QwhCkyAtRT/x6cTv7Eg9ipbUnO6YD/h7OTBnaHpMaKPDX2DWy4OlRQZhe9UPJdeZk1hl+Pr+lxrYvhNAvKfJC1AM7L+9j26VdWNOY7OhgWjk78eTIQF13tTXJ3cWW/wzzp/h8EKoSW3Zd2c+h5CM1vh8hRO2TIi9EHXcw6S9+Ov8bViobsqOCcbV14NmxHbC2NKu1fYa1deb+rr4UnQ7BRGvBd2d+5ELOpVrbnxCidkiRF6IOO5YaxXdnfsRSZU1OdAiNLex5LiKYxo0san3fw7u3IrRVK4rOdKBcq2X5idVkFmXX+n6FEDVHirwQdVRsxim+Pvk95ipz1HEhWClNmDU2mKb21nrZv4lKxZSh7Wlh5UlpQnvyy9QsPfE1xZoSvexfCHH3pMgLUQedy47ny9g1mGBC8ZkwKLLn6VFBtHS5dXe1NcnKwoynRgVine9DeVpLktQprD61TlrcC1FPSJEXoo65nJfIFzErKddq0caHUprbhKn3+dHWw8Eg8TjbW/PkiADKr/iB2ono9Fi2XNxhkFiEELdHr0U+OjqaiRMnXjf9119/ZcyYMYwbN47XXnsNrbbiLGHEiBFMnDiRiRMnMmfOHH2GKoRBpBSksjj6S0rKSzFPDqMg3ZGJg9oS1tbFoHG183RgfP+2FJ3tgEmZDVsv7eJYapRBYxJC3FrtNc/9l+XLl7Np0yasrSvfTywuLubjjz9m8+bNWFtbM2vWLPbs2UOPHj0AWLNmjb5CFMKgMoqyWBS5nIKyQmzSw8hIdGJETy/6BLsZOjQA+oa4cSVNzf5TpTQK/Is1p9bT1NoJz8YtDR2aEOIm9HYm7+HhwaJFi66bbmFhwffff68r/hqNBktLS06fPk1RURGTJ09m0qRJREVF6StUIfQutySPRVHLyS3Nwz4vmIyLzvQNdWN4t1aGDk1HpVIxYaAvbZq6UXQ2kDKthqUxq8gpyTV0aEKIm1ApiqLoa2eJiYnMmjWL9evX33D+mjVr2LdvH8uXL+fs2bNER0czZswYLl26xGOPPca2bdswM6v64oNGU46ZmWlthC9ErcgvUTN394dcyUvBtawDlyKb0z2oBS9MDMfUpOZ6s6spOfklzPpkH9mWJzH3OENrx1bM7TsTC7Paf6xPCHF79Ha5viparZb33nuPixcvsmjRIlQqFV5eXnh6eur+b29vT3p6Os2bN69yW9nZhdXer7OzHenp+XcbvrhNkvf/V6wp5tOo5RUFvtyPS5HNaOdhz6R7fMnKVNfovmoy79NHBDB/TTGKrZrzXOKTA1/zsN+4GulD3xjJMW8YDSXvzs52N51XJ1rXv/baa5SUlPD555/rLttv3LiRhQsXApCamoparcbZ2dmQYQpRo8rKy1gas4qEvCs0V/ly6VhLPFzseGpUEOZmdeKjeVMernZMHupH8QU/TIocOZIayY6EvYYOSwjxLwY7k9+8eTOFhYUEBASwceNGwsPDefjhhwGYNGkSo0ePZs6cOYwfPx6VSsX8+fNvealeiPqiXFvOV3HfcDbnAi3MfbhwsBXO9tbMrOXuamtSp/auJKZ78+sRDbZBf7EpfhvNbFwIcvY3dGhCiP/R6z15fbidSzMN5VJOXdPQ865VtKw6+T1HU6NoYelJ/IG22FlbMmdiGK4OjWptv7WRd62isPiHE0Qnx9PI/2/MzUx4Lmw6brZV31ZraBr6MW8oDSXvdf5yvRANhaIorD/7C0dTo2hm6UbCYV8szM2ZOTa4Vgt8bTFRqXhsuB/NGzWn6HwAJeWlLI35mvzSmm1PIIS4M1LkhdCjTfHb+CPpEC6WrqQc9QetGU89EIhns5v/Eq/rrC3NeHpUINZF7miSWpNZnM3yE2vQaDWGDk2IBk+KvBB6sj1hD9sT9uBo6UhWdAdKikz4zzA//Fo5Gjq0u+bi0IgnRgRQntwaVW4LLuReZN2ZnzGyu4FC1DtS5IXQgz+SDvHLha00sWhCyamO5OeZ8OBAXzq1dzV0aDXGv5UjEf3aUHjOH7NSe/5M+Zu9iQcNHZYQDZoUeSFq2ZGrkaw78zM25jaYXOxCRoaKYd1a0T/M3dCh1bgB4e708HdHfbIDZlorfji3mVOZZw0dlhANlhR5IWrRiYyTrD61DktTSxpf7UlykopeHZozsqeXoUOrFSqViomD2uLt7Ir6VDAqTPgq7htSC9IMHZoQDZIUeSFqydns83wZ+w1mKlOa5/Uh/gKEtGnKxEFtjbpnOHMzE2aMDMTexJXiC34UaYr54sTXFJZVvzdKIUTNkCIvRC24lHeZL2K+RlEUfMr6cTIOfN2bMPU+f0xNjP9j18TWkhkPBGKS2xLSvEkrzGBF3LeUa8sNHZoQDYrxf9sIoWfJ6qt8HrWC0vIyAs0GcPyYCndnG54eHYSFecMZPMmreWMevbcdRZfaYKZuxqmss/x0/jdDhyVEgyJFXogalF6YyeKo5RRoCuloM4BDB1U4NbZi5thgGlmZGzo8vevi34x7O3uSfzoAc00T9iQe4GDyX4YOS4gGQ4q8EDUkpyT3f2PC59OlST/27zXF1tqcWREdcLCzNHR4BjOqtw9BXq7kx3XADEvWnfmZc9nxhg5LiAZBirwQNUBdWsCiyOVkFmfRxbEnB/ZYYWFmyrNjOtDcycbQ4RmUiYmKx4f742rTlIJTQWgVhS9j15BRlGXo0IQwelLkhbhLRZpiPov+kquFaXR06szhPY3RahWmjwzAu0VjQ4dXJzSyMuPp0UFYlrpSluCHuqyApTFfU6wpNnRoQhg1KfJC3IXS8jK+iFnJ5fwkQp1CifrDhaKSciYPbU+At5Ohw6tTmjk24on7/dGkuWOS5UVywVW+Pvk9WkVr6NCEMFpS5IW4Qxqthi9j13A+5yKBjgGc+8uTPHUZ4/q1pqt/M0OHVycFejsxpk9rCs63waLYlRMZJ9kc/7uhwxLCaEmRF+IOaBUtq0+uIy7zNO0cfEmLbktaVhH3dvbgnk4ehg6vThvUqSVd/ZuTGxeApbYx2xP28PfV44YOSwijJEVeiNukKArfn/mRY2nReDduRcnZYC6lFNA9oBmj+/gYOrw6T6VS8fDgdni5OJIbG4QZFqw9vZFLeZcNHZoQRkeKvBC3QVEUfrrwGweT/8bdtgWNUrpy8mIeQT5OPHxvO6PurrYmWZibMuOBIBqbOVJ4JpBybTnLYlaRU5Jr6NCEMCpS5IW4Db8n7GHX5f24NnLGLb8fR+Ky8WnRmGkjAjAzlY/T7XCws2TGyEBUaheUpPbkluazNGYVpeWlhg5NCKMh30pCVNPexINsjt+Go5UDAcoQ9h7JoLlTI54Z0wHLBtRdbU3ycWvCw4PbUZTUEos8Ty7nJ/LNqQ0oimLo0IQwClUW+V27dlFWVlblBgoKCnj33XdrNCgh6pq/Uo6x4ewv2FnY0s36fn7dl4qDnSXPRQRja93wuqutSd0Dm3NPRw9yz7TFqsyZY2nRbLu029BhCWEUqizyM2bMIC8vr9K0Pn36kJSUpHtdVFTEypUrayc6IeqA6PRYvjm9gUZm1gxyHMMP21OxsTJjVkQwjo2tDB2eURjT1wd/TyeyTwRgiS2/XvydqPRYQ4clRL1XZZG/0SWz3NxctFrpvEI0DKezzrEidi1mJmbc1yKCdVtSMTNV8cyYDrg1bdjd1dYkUxMTnhgRgIudPbkngjBTmbMq7juu5CcbOjQh6jW5Jy/ETcTnJrD0xCoARnlEsP7XDDTlCk+MCKC1WxMDR2d8bKzMeXpUEJblDpScD6RUW8bSmK/JL1UbOjQh6i0p8kLcQJI6hc+jV6DRahjrPZaftuRRUKzhkXvbEdy6qaHDM1otmtrw+H3+lGW6YJbWjuySHJadWE2ZVmPo0ISol6TIC/EvaYXpLIpaTpGmiDE+o9i6o5Ts/BJG9/GhR1BzQ4dn9IJbN+WB3t7kX/LEutCD+NxLfH/mR2lxL8QdMLvVAr/++is2Nv9/71Gr1bJ161YcHR0BUKvlUpowHtnFOXwauZz8UjUP+NzHH/tUJGcUcE/HltzbWbqr1ZchXTy5kqbm7zgtTmGFHE45SgubZvT36GXo0ISoV6os8i1atGDVqlWVpjk5OfH9999Xmta8uZzdiPovv1TNoqjlZJfkMLTVIE78bceFpEy6+Lkytl9r6c1Oj1QqFY8OaU9qVhEJ0QE4hB7hp/O/0czGBX+ndoYOT4h6o8oiv3u3PKsqGobCsiI+i/qS1MJ0BrTsTfKpZsRcSMXfy5HJQ9tjIgVe7yzNTXlqVCBvfn2EvLggrP2PsCL2W14In04zG1dDhydEvXBH9+RLS0uJi4vj6tWrt7VedHQ0EydOvG767t27GTVqFBEREaxfvx6ouC3w2muvERERwcSJE0lISLiTUIW4pdLyUpbErOSKOpnuLTqjSWzLnydS8Wpux/SR0l2tITk2tmL6A4FQaI82IZDi8mKWxHxNQVmhoUMTol645bfX6tWrGTJkCImJiQDExcUxYMAARo0aRd++fXnuuecoLb11X9PLly/n1VdfpaSkpNL0srIyFixYwIoVK1izZg3r1q0jPT2dnTt3Ulpayrp163juuedYuHDhHb5FIW5Oo9Ww7MRq4nMvEebSAae8cLb+dRlXx4ruaq0sbtlsRdSyNu72TBzUlsKrrlhltyWjKJOvYr+hXFtu6NCEqPOqLPLfffcdH330EYMHD8be3h5FUXjuuedQqVRs3ryZPXv2kJKSwpIlS265Iw8PDxYtWnTd9AsXLuDh4UGTJk2wsLAgLCyMo0ePcuzYMXr27AlAcHAwsbHS+5WoWVpFy9dx33Eq6ywBTu1oSx/W7b5AE1sLnhvbgcaNLAwdovifXh1a0D/UnexzrbAtdedM9nk2ntts6LCEqPOqPE1Zt24dr7/+OiNGjADg6NGjXLp0idmzZ9OmTRsAnnzySV5//XWeeeaZKnc0aNAg3dWAf1Kr1djZ2ele29jYoFarUavV2Nra6qabmpqi0WgwM6v6zMrBoRFmZtUfLMTZ2e7WC4kaZ+i8axUtS4+sJTL9BH7ObRjUfDQLVhzDxsqMt6Z2w6uFcXZ2Y+i8340Z40JIzysmJqYdLp2K2J/0J77NPLmndf1ocV+fc1+fNfS8V1kxL168SHh4uO71n3/+iUqlok+fPrppXl5epKWl3XEAtra2FBQU6F4XFBRgZ2d33XStVnvLAg+QnV39e3XOznakp+ffXsDirhk674qi8MP5zey58icedu70dxjBe6siUalUzHggEFtzE6M8Lgyd95rwn6HtefNrNenH/WkSeoQVx9dho7XD16G1oUOrkjHkvj5qKHmv6odMlZfrraysKCz8/6L5559/4u7uTqtWrXTTUlJSaNLkzs96fHx8SEhIICcnh9LSUo4ePUpISAihoaHs378fgKioKHx9fe94H0L809ZLO9lz5QDNbFwZ4zmez384TammnCfu96eth4OhwxNVsLWu6PrWQrGj8HQHAL488Q3phZkGjkyIuqnKIt+tWzfWrl0LwPHjx4mOjmbIkCG6+VqtluXLl1c626+uzZs3s27dOszNzZk9ezZTpkxh3LhxjBo1CldXVwYOHIiFhQXjxo1jwYIFzJkz57b3IcS/7blygN8u7sDJypFJrSexZOM51EVlTBrUllBfZ0OHJ6rB3cWW/wzzozTHHrPkIAo0hXxx4muKNMWGDk3UEerSAlYc/4nXty6lrLzq4dKNnUqpoq/IK1eu8PDDD5OXl0dRURHe3t5899132Nra8uuvv7J06VLS0tL47rvv8Pb21mfcN3U7l2YayqWcusZQeT+UfIRvTm+giYUd0wIeZ/kPl0hML2BkTy+Gd/fSezz6ZmzH+6YDF/n5wEVc/OLJt61oPDk16BFMVHXvkUdjy31dVVhWyMaTO/k7/TCKScV4B60b+fFs54eNujOrqi7XV3mTu2XLlmzdupWDBw9iYmJCt27dsLCoaHFcVFRE586defjhh2nZsmXNRixEDYtMO8Ha0xuxMWvE1IAprP01kcT0AvqFujGsWytDhyfuwLDurbiSrubYSS3NwguIzTzNLxe2MrL1UEOHJvSsSFPEpjN7OHD1IFpVGUq5BU4FIWSo4jnPSbZe3M0Q7/6GDtMgbtmSzdLSkn79+l03fcyYMbUSkBA17WTmGVbGfYuFqTnTgiazaVcGZxNzCW/nwoMDfI36F74xM1GpmDK0ouvbxMh2uHQsZOflfbSwaUbn5mGGDk/oQbGmmK0X9rEn8Q/KVaUoGnOaFAQxIXQgAa1c+emv0+zIXstvl36nZePmBDb1M3TIeldlkf/kk0+qvaFbPUInhCFcyLnEshOrUalUTA18hP2HC4k8l0F7TwceG+aHiYkU+PrMysKMp0cF8uaqo2RGB2LX4W++Pb0R50ZN8W7iaejwRC0p1pSw89IBdlzei4YSlHJzrHP9GRs4gE5tW+h+uE+6J4QDH8VT1HL//7pEnkEL22YGjl6/qizyS5YswcTEhPbt22NjY3PToR7lTEjURVfyk1gSs4JypZzHAydxMtaEfVHJeLjaMuOBQMzN6t69W3H7mtpbM31kAO9/H0XZhWDw+ptlJ1bxUvjTOFjZGzo8UYNKy0vZc/kg2y7upZQiFI0Z5lntuL9tX3oP8MTUpPJn2srCjId7d2LRrmxUraNZErOSl8KfxtbC5iZ7MD5VFvnXX3+dXbt2ERkZSceOHenfvz/9+/fXDTNr7Io0xRSUFdLUumG8X2OSWpDG4qgvKdaU8IjfONISGrP5z7O42Fszc2ww1pbSXa0xaevhwIMD2rBm+1maNg4i3ymKpTFfMzPsSSxNpefC+q60vIz9iYfYGr+bYqUQRWOGSYYvg7x7MbivDxbmN+8ALcinKSExQUQlqclyu8CXsWuYEfwfzEwaxndAla3rr1Gr1ezfv59du3Zx4MAB2rRpw4ABAxgwYADu7u76iLPaarJ1/cq4bzmaGkWzRi6EuAQS4hJEC5tmcuXiLtV2S+Os4mw+PLaE7JIcxrV9AKt8b774ORa7Rua8PDEMF4dGtbbvuqwhtPBeve00e6OSaBEcT7bFOUKcA5kcMMHgLe4bQu5rQ1l5GQeS/+K3C7so0haglJuipHnRq0UP7uvaBhsr8yrXv5b37PwSXl5+CJNWx8H+Kt1bdGZ82weM5ru8qtb11Sry/6TRaDh06BC7d+9m79692NvbM2DAAKZPn37XgdaEmizyCXlX+P3Sbk5mnaFMW/E4hrO1EyEuQQQ7B+Bh5240B4k+1eYXXl5pPh8dW0JaUQYjfIbgrgTx0YZozExNeOnBUDybNdwuLhtCodGUa3n/u0jOJmXTotMJspUUhngNZKjXQIPG1RByX5PKtBoOJR/h1ws7KSjPRyk3pTy1FR2dOjOqR3sc7CyrtZ1/5n3HkSt8t+cUjqHHKTLJYozv/fRx716bb0NvarTIQ0UnOMeOHWPXrl1s2LABrVZLZGTkXQVZU2rjOfliTQlxmaeJTD9BXMYpSrUVnSs4WjkQ4hxIsEsgrRq3NPjZQn1RW194hWWFfBy5lCR1Cvd49iXYpgfvfHscTbmWmWM60L5Vw77t0lAKTV5BKW+tOkJmYT7OHY+hLs9lSsBDhLoEGSymhpL7u6XRajiccpRf43eSX5aHUm6KJtUDf9twxvb0o7nT7d1L/2fey7Va3lp1lCvZ6TiEHaFEW8yTHSbT3rH+96ZaI0X+2iX7PXv2sH//fszMzOjTpw/9+vWjR48eWFpW75dVbavtznBKy0s5mXWWqLQTnMg4SXF5xdC59pZN6OAcQIhzID72raTgV6E2vvCKNSUsjvqSi3kJ9HLrSh/ne1iwNpL8glKeGBFAx3YuNbq/+qghFZrLqfnM/+YYKms1Vn6HUVCYFTYNDzvD3F5sSLm/E+Xacv66epzf4neQU5qDojVBk+qBl2kwEb0C8G7R+I62+++8xyfn8fbqozRtUURRywNYmFrwQvgMXBvV794u77jIJyYmsmfPHnbv3s3Ro0dxc3OjX79+9O/fn9DQ0Dp5qVqfPd6VaTWcyTpHZNoJYjLiKNQUAWBnbksHl4qC38beG1OT6o+K1xDU9BdemVbDF9ErOZ19jo6uodzvOYKF3xwnPaeYh+7xpV9o3Wo3YigNrdAcOZ3Gkp9jcXDLpsTtb5pYNubF8KdpYqn/WzYNLffVVa4t50hqJL/F7yCrJBtFa0J5WktcygIY2zOAAC/Hu6ozN8r7mu1n2HM8iU7dSzlRthvXRs48HzaDRubWd/t2DOaOi3z79u0xMzPTtayvquvarl273l2UNcRQ3dqWa8s5m32ByPQYotPjUJdVjKBnY96IDk39CXYJpK1D6wbTorMqNZ33r+LWEp0eS2BTPx5qM573v4/mcqqa4d1aMbJX3ehuuS5oiIXmx/3x/PrnJVq0TyHbLhqvxh48EzIVc9OqG2zVtIaY+6poFS1HU6P4LX4HGcWZFcU93R07dXtGd/Onk58rJjVwEnmjvBcWl/Hy8r8oKtHQc3AOh9L+pL2jL9OCHq23J2R3XOTbtWtXrR2oVCpOnTp1+5HVgrrQd325tpwLuReJTIslKv0EeaUV+7A2syKwqR8hzoG0d/TV+xdNXVFTedcqWr45tYG/rh7D196HxwIe4bMfTnIqIZvewS2YNKhtnbzaZCgNsdBoFYXPfjxB5Ll0PMIvkG5ynk7NQpnUPkKvx0ZDzP2NaBUtx9Ni2HJxB6mF6aBVoUl3xzKnLfd38qN3cAvMTGvuVufN8n447irLNp8kwNsBK99I4rJO07dlD0a3ua/G9q1PNd7wri6rC0X+n7SKlou5l4lMjyEqLZbskhwALE0tCHBqT4hLEH5ObRvUs7w1kXdFUdhwbhP7Eg/SqrEH04P+w9dbznP0dBqhvs48OSJAerP7l4ZaaIpKNMxfc4ykzDzcu54gU3OVET5DGOjZR28xNNTcX6NVtESlx7Ll4g5SClJBUaFJd8Mkow2Dg9txT8eWtdJ3xc3yrigK738fxamEbP5zXxt25a3jamEaE9qNpluLTjUeR22TIn8T+v7gKYpCQv4VItNOEJV2goziLADMTczxd2pHiHMA/k3bY21mpbeYDKEm8v5r/O9svbSLFjbNeCZkKj/vTWT38SR8W9rzXEQHzM3q52W32tSQC01adiFvrTpKsVKAY9hRCjRqpgY9rLe+zBtq7hVFITojji3xO0gqSKko7hkt0Ka0pq9/G4Z1a0Vjm9o7wakq71ezCnntq7+wtTZn5kNtWBSzhOLyEp4OeZzW9vVrVEop8jdhyA+eoigkqpOJSjtBZPqJiktXgJmJGe0d2xDiHERg0/Y0Mje+jlvuNu87L+/jp/O/0dTaiVmh09h/NJOf/riIu7MNsyeE0ugWHWQ0VA210Fxz8lIWH66LxsaxAFWbQ5ioVDwfpp++zBta7hVFITbzFL/Fb+eKOrmiuGc2pzzJh86tvRnR0xtn+9pv6HarvP/8RzybDl5iYHhLQkNVLI7+kkZm1rwY/hRO9ainUynyN1FXPniKopBSkEpkesUZfnLBVQBMVaa0dWhNiEsgQU39jaa/5bvJ+8Hkv/j29A/YWzZhVug0Ys8UsWrbGZwaW/HyxLBqd5LRENWV492Qdhy9wnc7z+HqnUNe08M0tXLkhfCnav2z1VByrygKcZmn+e3iDi7nJ4ICmszmaJJ9CGjhyaje3ni46u/phlvlvUxTzn+/+pv0nCJee7gjCZpY1p39iRY2zXgubDpWZvXj+0SK/E3U1Q9eakEakekVjfau5CcBYKIyoY29N8HOgXRwDjDIY0A15U7zfiw1mpVx32Jj3oiZoU+QnGTCZz+dwMaqorvaZo7Gd9WjJtXV412fFEVh5dbTHIhJwTM4hTSLaNrYe9d6X+bGnntFUTiddY5fL27nUt5lAMqzmlGW2BovxxaM6eNDWw8HvcdVnbzHXczig3VReDW345WJ4Ww49zP7kw4R1NSfxwIn1os+T6oq8vI8Vx3kauPCYJt+DG7Vj4yizIp7+OmxnMk+z5ns86w/+zM+9q0Idg4k2DmgQYy0FZtxiq9PfoelqSXTg6eQm2nBF79EY2FmysyxHaTAi2pRqVRMvKctVzMLOR+l4NVVzbmcC2w4+wvjjKgvc31RFIUz2ef57eIO4nMvVUzLbkZJog/NGrnywL0+hPo2rdN59fdypLOfK3+dTGVfVBKjg+/jakEaMRlx/Bq/nft8Bhs6xLsiZ/L16Nd1dnEOUemxRKbFEJ+bgELFn86rsQfBLoGEOAfWi/tIt5v3c9nxfBb9JQAzgh/DstSZhWuPU1pWzjNjggjwcqqtUI1KfTvea1OuuoQ3Vx0lp6AAt64xZJalMdZ3BL3du9XK/owx9+eyL/Drxe2cz7lYMSHXleIrPtibOnN/Dy+6Bza7buhXfatu3nPVJby8/C8A5j/WGVNLDe8dXUxGUSaP+I2nY7OQ2g71rsjl+puozx+83JI8otNjiUyP5Vz2BV3B97BzI9g5kBCXQFzqaFeNt5P3y3mJfBK5lFJtGU8EPYKrqSdvf3OMXHUpjw/3o4t/7TeaMhb1+XivDZeu5rHgm+OYWRVjE/QXReVFTO8whXaObWp8X8aU+/M5F/nt4g7OZp8HwCTflcIEb6zLnRja1ZP+Ye5VDv2qT7eT993HE/lm+1m6+Lny+H3+XC1I5b2jn6FRNMwMfYJWjT1qOdo7J0X+Jozlg5dfqiYmPY7I9BOcyT6PVtEC4GbbnGDnAEJcgmhu42rgKP9fdfN+tSCVj45/QUFZIY/6P0gbu/Ys+OY4qVmFjOvfhns6ttRDtMbDWI73mnT45FWWbTqJU/NCSjwOYvm/vsxr+geyMeQ+PjeB3+K3czr7HABmha6oL3phVuLIgDB3hnT1vOXQr/p2O3nXahXeXnOUiyn5PDcuGP9WjsRmnOKLmK9pbGHLix2fxt6ySS1HfGekyN+EMXzw/q2grJATGSeJTDvB6ayzaJRyAFwbuRDyv0v6brbNDXqPrDp5zyjK4sNjn5NbmseEdqMJbRrKe99FcjEln3u7eDCmT2s9RWs8jPF4rwkb9p5n6+HLeLTLIb3x4Vrpy7w+5/5S3mV+i9/ByawzAFgUu5If3wqlwIGeQc25r7sXjo3rZt8et5v3hKv5vLnqCC721rw5pRPmZqa6R3Y97NyZGToNizrYU6kU+Zuozx+86ijSFBObcYqo9BPEZZ6mTKsBoKm1EyH/u6TvYeeu94J/q7znluTx4fElZBRl8kDrYfR268EnG2OIu5hFj8DmPDqkXZ1uyFNXGfvxfqe0WoVPf4gh5kImbTolk0gMfo5tmdbh0RprWV0fc385P5Hf4ncQm1nRZbl1mSs55zzRqh0J83Xmgd7etz30q77dSd6/3XGWnccSub+HF/f38EJRFL45tYHDV48S5tKBR/0frHPfP1Lkb6I+fvDuVLGmhJNZZyqGyM08RWl5KQAOlvYVZ/gugbRq7KGXx0WqyntBWSEfH/+C5IKr3NuqP0O87uHLzSc5fDKVDj5OzBgVaPDGPPVVQzreb1dhsYa31xwlJbMAnx5nSS69SL+WPRnVZniNbL8+5f5KfjJbLu4gJiMOAJtyV7LPelCe70TblvaM7uODj1vdvGz9b3eS96ISDS8vP0xBkYa3pnTC1bERZVoNnxxfysW8BIZ7D2Jwq/61FPGdkSJ/E/Xpg1eTSsvLOJV1lsi0E5zIOElxeTEATSwaE/y/IXJ97L1qreDfLO/FmmI+jVpOQt4Vert3Z3Tr4azbfYEdR6/g49aY58eFYFlHGvTURw31eK+uq1kVXd+WaUtw7RxJVmkGE9qNoVuLjne97fqQ+2T1VX67uIOo9BMA2CkuZJ/zpCzHkZYudozq7UOg990N/apvd5r3a8MU+7dyYFZEMCqVirzSfN49sojskhweC5xEsHNALUR8Z6TI30R9+ODVtjKthjNZ54hMP8GJ9JMUaAoBsDO3pYNzxRC5vvY+NToE443yXlZexufRKzibc4HOzcJ4qP0Yfv/rChv2XqBF04ruam2t6969sPpEjvdbi43P5KMN0djZl2LW7hCl2tIa6cu8Luf+akEqWy7u5HhaDAoKTVQu5JxvRXGmA02bWDOylzeda2joV32707wrisJHG6KJjc9i6n3+dParaLh8JT+ZD499BioVz4dNx822eU2HfEekyN9EXf7gGUK5tpyzOReI+l/nO+qyAgBszBoR6FwxRG5bxzaY32XPYP/Oe7m2nOWxaziRcZIOzgFM8Z/Aodg0Vmw5hWNjS15+KKzONuypT+R4r55tf11m/Z7ztGhVRK7LHzQyt+bF8Kdxsr7zHtvqYu5TC9LYcmknx1KjUVCwN3UhP94Ldao9do0sGNatFX2C3TA3q7+3x+4m72nZhfz3q79pZGnG24911o2JEZl2gi9j1+Bo5cCL4U9hZ2FbkyHfESnyN1EXP3h1hVbRciHnoq4//dzSijxZmVoR2NSPEJdA2jv63lFL03/mXatoWX1yHUdSI2nn0IYnOjxKXHwOi384gbWlKXMeCqNF07rduKe+kOO9ehRF4ctfT3Eo7iptgnNItDiMm21zZoU+ecd9mdel3KcVZrDt0i7+vnocBQUHM2eKErzJTrLH0sKMQR1bMqiTR60M/apvd5v3zX9e4qf98fQLdeOhe9rqpm+9uJNfL27Hp0krngp5/K5PfO6WdGsrbpuJyoQ2Dj60cfBhdJv7uJR3mci0E0SmneBI6nGOpB7HwtSCQKf2BLsE4u/UDkvT2xsyUlEUNpz9hSOpkXg19uTxoIdJSC7gi59jMTNV8cyYDlLghd6pVCoeubctV7MKORcF7bsHckl9gtUnv+c/9aQv8xvJKMpk6/+Ku1bR4mjeFE1SG5IvNcbUxIT+YW4Mr+WhX+ubwZ08OBx3lT3Hk+ge2Byv5o0rprfqT3LBVY6nxbDuzE9MaDe6zrZVkDP5OvLrur5QFIXL+YkVBT/9BBlFmQCYm5jj79SWYOdAApq2x9rs5pfXr+X9lwtb2Z6wBzfb5jwb8gTZuVoWfnOMopJynhoVSIfWTfX1thoEOd5vT3Z+CW+uOkJeYTHePU6TXHyZwa36M9x70G1vy5C5zyzK5veEXRxKOYpW0eJk0RSTtLZcPmsLqOji58qIXt646GHoV32ribyfTsjm3e8i8XC15b8Ph+ue7iktL+XD40u4kp/EqDbD6deyZ02EfEfqxOV6rVbL3LlzOXPmDBYWFsybNw9PT08A0tPTmTVrlm7ZU6dO8dxzzzF+/HhGjBiBnV3FG3B3d2fBggVV7keKvP4oikKiOoWo9Ioz/NTCNADMVKa0c/T93xC5fjQyrzx4jLOzHWuPbuKXC1txsW7KzLBpaIrNeXvNMbLzS5gytD3dA+tGgxZjIsf77buQnMs7ayMxt9TgEHqE7NJsHvV/kHDX4NvajiFyn12cw7aE3RxKPkK5Uo6TpRPW2e05F2sDqAjwdmR0bx+9Dv2qbzWV9y9/PcmfsVcZP6ANA8P/v6fN7OIc3j26iPxSNdM6TMbfqW0VW6k9daLIb9++nd27d7Nw4UKioqJYunQpS5YsuW65yMhIPvroI1auXIlGoyEiIoKff/652vuRIm84KQWpRKbFEJUeS5I6Bai47N/WoTUhzoEEOftjZ2FLVG4Uy499i4OlPbPCpmGh2LLgm2OkZBYypq8P93b2NPA7MU5yvN+ZgydS+Oq3U7g001Dq9QdapZyZodPwbFz9bpX1mfuckly2J+zhYNJfaJRynCwdsS8I4GRkI7QKeDVvzJg+PrTz1P/Qr/pWU3nPKyjlleWHKdcqvP1YFxzs/r9txsXcy3wc+QVmKjNeCJ9BMxuXu97f7aoTRX7BggUEBQUxdOhQAHr27Mkff/xRaRlFURg1ahTvv/8+3t7eREdH8+KLL+Lm5oZGo2HWrFkEBwdXuR8p8nVDamH6/1rpn+ByfhIAKlR4NfHgYu5lbMwbMSt0GvYWTrz/fSQXkvK4p2NLIvq1rrP3tuo7Od7v3Pe7zrH9yBV82hWT0ngfjS3seLHjU9Xuy1wfuc8tyWdHwh7+SD6MRqvB0dIB19IOxB6zolQDro6NGN3bm1Bf5wbzGavJvO+LSmLVtjOEt3PhyRGVn5H/++pxVp38HmdrJ14Ifwobc/0OfV0nGt6p1Wpsbf//UQNTU1M0Gg1mZv8fwu7du2nTpg3e3t4AWFlZMWXKFMaMGcOlS5d47LHH2LZtW6V1/s3BoRFmZtV/pruq5Ig754wdAZ7ePMT9pKkz+CsxisOJxzmXeZFG5ta81vcZ3Bu78fbKv7mQlEefUHemjw3BxKRhfPkYihzvd+bJMcGk5xYTeTqd0B7dOFV6kBWnvuGNvrOwMKteQ7Xayn1ucR6/nNrO9gv7KS0vo2kjR7xMwzl+yJykwnIcG1vx4KC2DOjogalp/Ww0eDdqKu8P9G/LX6fSOHo6jYSMQsLb//+gX0Ode5OjzeKX09tZc/Y75vR6CrMa7FvkbuityNva2lJQUKB7rdVqryvWmzZtYtKkSbrXXl5eeHp6olKp8PLywt7envT0dJo3v/n92uzswmrHJGc2+qHCki5Oneni1JmcklycnewozVfx/uojHD2VSoCXIw/2b01mptrQoRo1Od7vzuQh7ZiXrub4AYWA3v5cyIrj4wMrecRv/C3PjGsj9+rSAnZe3se+xIOUasuwt2xCkFUYscds2J9XRiNLFaP7+NA/zB1Lc1OysgpuvVEjU9N5H9+/DW+sPMJnG6KY95/OlYbUHdC8H/EZVziReoqlh75lrO+IGtvvrVT1Q0ZvP+tCQ0PZv38/AFFRUfj6+l63TFxcHKGhobrXGzduZOHChQCkpqaiVqtxdq6bY6SL6rG3bIK9dRM27rvAwdireDVvzJMjAzBrgGcYon6xsTLnqVFBWFmYceZPd1pYu3M0NYrtCXv0GkdBWSGbLmzjtUML2HF5L9Zm1nSz7w+n+vLHHgvyC8oZ3NmDhU90ZUgXT+kKuga1dLFlYEd3MnKL+fXQpUrzTFQmPOI3nhY2zdiX+Cd/JB02TJD/orcz+YEDB3Lw4EHGjRuHoijMnz+fzZs3U1hYSEREBFlZWdjY2FT6RTx69GjmzJnD+PEVv5Tnz59f5aV6UT/8vO8CWw9fppljI54dU/GlKUR90KKpDVPv8+fTjTFkRvvTJCCfTfHbaGbjSgdn/1rdd2FZIbuv/MGeKwcoLi+hsYUdXRx7cy6qCbuS1KhUxfQMas79Peru0K/G4P4eXvx9Ko2thy/T1b9ZpZH4rMysmBr0CO8e/ZT1Z3/GtZEzvg4+BoxWnpOXy5d6VKYpZ29kMt/tOoe9rQUvTwyjaRPjeza3rpLjveb8dugSP+yLx6OVltxme27Zl/nd5L5IU8SeKwfYfeUPijTF2Jnb0smpKwlxjpy4kAtASJumjOrtI51H/UttHfPHz6az+McTtPOw54XxIdfdrjmXfYFPo5ZjbWrFix2foqm1U43H8E91ouGdaLhSswvZF5nMgRMpqIvKsLEyY9bYYCnwot4a0sWTxPQC/jqZSoB9Dy5Y7OGLmK9rtC/zYk0xexMPsuvyfgo1Rdia2zDQbSCp51z47WAmCrn4/m/o19b1ZOhXYxHSpinBrZsSdT6DQ3FX6RZQ+cddGwcfxvmO5NszP/BFzNc8Fza9yg7CapMUeVEryrVaos5lsjcykbhL2QDYWptzbxcPRvdvi6q83MARCnHnKrq+bcfVzEJioyC0Z2dOFf/F8hNreDrkMczuoi/zYk0J+5P+ZOflfRSUFWJj1ojBLQeSd9mNLZtS0ZRn4u5sw+g+PgR6OzWYx+HqEpVKxYMD23AyIYt1u88T5NP0ulEyu7t1JqngKvsSD/J13HdMDXrYIF0iy+V6uXxZo7Lyitkfncz+6GRy1KUA+Lo3oU+oG2G+LpibmUjeDUTyXvOy8op5c9VR1IWl+Pe5xPmC03Rr3pEH/9WXeXVyX1peyv6kQ+xI2Iu6rABrM2v6tOhB2VUPdh65SlFJOU6NrRjZy4sufs3kcdNqqO1jfuvhBDbsvUDv4BY8PLjddfPLteV8Hr2C09nnGOjRhxGth9RKHHK5XtQqraJw8mIWeyKTiD6fiVZRsLY0pX+oO31CWuDmbPihGIWoDY6NrZg+MoB3v43kwmEvmnfK5c+UI7SwbU7flj2qtY3S8jIOJB9me8Ie8kvVWJlaMdhzABbZPmzdmkJeQRK21uaM7+9Nn5D6PfSrsRnYsSV/xl5lX1Qy3QOa09q98m0TUxNTpgRM4L2ji9lxeS/NbVzp3DxMrzHKmbyc2dyxvMJSDsaksDcqifScYgA8Xe3oG+pG5/auWFrc+NEdybthSN5rz/7oZL7eeprmrirKWx9AXabmyQ6T8ftfX+Y3yn1ZeRkHk/9me8JuckvzsTS1oI97DxyK2/PbgWTSsouwNDdlUCfjGfpV3/RxzJ+9ksPCtcdxd7bhtUc63vBx4NSCNN47tpiy8jKeDX0CryY123W3nMmLGqMoCucSc9kblcTR02loyhUszEzoEdScviFuuqEYhWhIenVowZU0NbuOJdLOoTtF9jtZEbeWF8Jm4PqvvszLtBoOJR/h94Td5JTkYmFqwT2efXFXgvh1fzIJqRcwNVHRL9SN4d29aCJDv9Zpvi3t6RnUnD9iUth5NJHBnT2uW8bVxoXJ/hP4PHoFS0+s4qXwp3GwstdLfFLkRbUUlWj4M/Yqe6OSSEqv6DmruVMj+gS70S2wGTZW5rfYghDGLaJfa5IzCjh1OpvwLr2I0+zmi5iveSF8BmBHubacwylH2XppF9klOZibmDPAoze+lqFsOXCVXxLOANDZz5WRPb1wcdBv/+fizo3p25rIcxn8cuAiHdu54NTk+pb0fk5teaDNMH44t5mN5zbxWOCkG2yp5kmRF1VKuJrPnsgk/jqZSklZOaYmKjq1d6FPsBttPeylZa8Q/2NmasK0EQG8teoIRw9Dx36diFX/zVexa+lT3IUNJ34lszgbcxMz+rXsSXDjzmw/lMbm06cACPByZFRvHzybyfgC9Y2ttTlj+7ZmxZZTfLvzLE+NCrrhcn3de6BCpbezeJB78nKP8gZKyso5ciqNPZFJXEzJA8CpsRV9QlrQI6jFXV8+lLwbhuRdPxLT1by95hiKoqVtr/Oczz8HgJmJGT1adKZz0+7sPZLJH9HJlGsVvJrbMbq3D+1bORo4cuOjz2NeURTe+TaSs1dyeGpUICFt9NcFu9yTF9WSklnA3shkDp5IobBEgwro4ONE31A3Aryc5JEdIarB3dmWx4f5sejHEyQd88W/iznujs6EN+nC4egc5m+OpbRMi6tjI0b18iasbcMZ+tWYqVQqJg5qy9wVf/PtjrP4eTretPGxPkmRb+A05VqOn01nb2QSpy/nANDYxoJhYZ706tBCeqUT4g6E+DozoqcXP/9xkaZxQTiFuLHwp5Ooi8poYmvBuP5e9AhsLgMzGRm3pjYM6uTBlsMJbPrzImP6tDZ0SFLkG6qM3CL2RSXzR0wKeQUVnda087Cnb6g7IW2aypePEHdpeLdWJKapOXomnbNXcrC2NGNUb28GhLeUkeGM2PDurfj7VCrb/75CV/9muBu4nxAp8g2IVqtwIj6TPZFJnLiQiQI0sjRjYHhL+oS0qDSakhDi7qhUKqYM9cPM9DRurnb0Dmp+XdenwvhYmpsyYaAvn2yMYc3vZ3hpQigmBrwdI0W+AchVl/BHTAr7opLJzKvotMa7RWP6BLvRqb0LFnJWIUStsLQw5fH7/KXRYwPToXVTwnydOXY2nYMxKfTs0MJgsUiRN1KKonDmcg57IpM4fjadcq2CpbkpvYNb0CfYTR7TEUKIWjR+QBtiL2Wxfs95gts0xa6RYTo1kiJvZAqKy/jzREWnNSmZhQC4OdvQN8SNLn7NaGQlf3IhhKhtjo2tGNnDi+93n2fDngtMHtreIHHIN74RUBSFiyn57I1M4u9TqZRqtJiZquji70rfEDdauzWRR3SEEELP+oe7czD2KgdOpNA9sBltPRz0HoMU+XqspLScwyevsjcymYTUivt9LvbW9A5pQY/A5ga7PCSEEAJMTUyYNLgt81cfY832s8x99MYD2NQmKfL1UGK6mr2RSRyKqxhj2kSlIqRNU/qGuuHXytGgLTmFEEL8P58WTegd4sbeyCR+//syQ7u20uv+pcjXE2UaLcfOVHQ1ey4xFwB7WwsGhrekV4cWODa+fkAEIYQQhjeqtzfHz6Sx+eAlOrV3xdlef52MSZGv49KyC3Wd1qiLygDw93KkT7AbwW2cMDWRTmuEEKIus7EyJ6J/G5ZvPsnaHWd5ZnSQ3tpJSZGvg8q1WqLPZ7I3MonYi1lAxShHgzt70Du4Ba4yBKUQQtQrXfxcORCTQsyFTI6fTSesrYte9itFvg7Jzi9hf3Qy+6OTyc4vAaC1exP6hrgR3tYZczPptEYIIeojlUrFQ/f48vqKv9kXlSxFvqHQKgqnLmWzNzKJyHMZaBUFKwtT+oa60TfYDXcXw/Z7LIQQomY0d7Jh9oQwrC31d8ImRd5A1EVlHIhJYW9UEmnZRQB4uNjSJ9SNLn6uWFnIn0YIIYyNd4vGet2fVBI9UhSFC0l57IlM5MjpdDTlWszNTOge2Iw+IW54N28sndYIIYSoMVLk9aCoRMPhuKvsiUwiMb0AgGaOjegT4ka3gGYyMpUQQohaIUW+Fl1Orehq9tDJVEpKyzE1URHezoW+wS1o5+kgZ+1CCCFqlRT5GlZaVs6R02nsjUziQnIeAE6NLRnSxZOeQc2xt7U0cIRCCCEaCr0Vea1Wy9y5czlz5gwWFhbMmzcPT09P3fyVK1eyceNGHB0dAXjjjTdo1apVlevUJVezCtkbmcTBEykUFGtQAUE+TvQJdiPIxwkTEzlrF0IIoV96K/I7d+6ktLSUdevWERUVxcKFC1myZIluflxcHO+88w4BAQG6adu3b69yHUPTlGuJOpfBnsgkTiVkA9C4kTlDu3rSq0MLvXZdKIQQQvyb3or8sWPH6NmzJwDBwcHExsZWmh8XF8eyZctIT0+nT58+TJ069ZbrGEpmbjH7opP5IzqZ3IJSANq2tKdvqBuhvs56H2VICCGEuBG9FXm1Wo2t7f937GJqaopGo8HMrCKEoUOH8uCDD2Jra8uMGTPYs2fPLde5EQeHRpjdRs9wzs521VquXKsQeSaNrX9e4uipq2gVsLEyY3hPb+7t2oqWrtXbjqhQ3byLmiV5NxzJvWE09Lzrrcjb2tpSUFCge63VanXFWlEUHn74YezsKv4YvXv35uTJk1WuczPZ2YXVjsnZ2Y709Pwql8krKOWPmGT2RSWTkVsMgFdzO/oEu9HJzxVL84ofFLfajvh/1cm7qHmSd8OR3BtGQ8l7VT9k9FbkQ0ND2bNnD0OGDCEqKgpfX1/dPLVazbBhw9iyZQuNGjXir7/+YtSoURQXF990HX3YcfQK63efp1yrYGFuQq8OzekT4karZvrtsUgIIYS4E3or8gMHDuTgwYOMGzcORVGYP38+mzdvprCwkIiICGbOnMmkSZOwsLCga9eu9O7dG61We906+lRcWk5LF1u6Bzanq78rjayk0xohhBD1h0pRFMXQQdSk27k001Au5dQ1knfDkLwbjuTeMBpK3qu6XC/NwIUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjJUVeCCGEMFJS5IUQQggjZaavHWm1WubOncuZM2ewsLBg3rx5eHp66ub/+uuvrFq1ClNTU3x9fZk7dy4mJiaMGDECOzs7ANzd3VmwYIG+QhZCCCHqNb0V+Z07d1JaWsq6deuIiopi4cKFLFmyBIDi4mI+/vhjNm/ejLW1NbNmzWLPnj306NEDgDVr1ugrTCGEEMJo6O1y/bFjx+jZsycAwcHBxMbG6uZZWFjw/fffY21tDYBGo8HS0pLTp09TVFTE5MmTmTRpElFRUfoKVwghhKj39HYmr1arsbW11b02NTVFo9FgZmaGiYkJTZs2BSrO2gsLC+nevTtnz55lypQpjBkzhkuXLvHYY4+xbds2zMxuHraDQyPMzEyrHZezs92dvylxxyTvhiF5NxzJvWE09Lzrrcjb2tpSUFCge63VaisVa61Wy3vvvcfFixdZtGgRKpUKLy8vPD09df+3t7cnPT2d5s2b33Q/2dmF1Y7J2dmO9PT8O3tD4o5J3g1D8m44knvDaCh5r+qHjN4u14eGhrJ//34AoqKi8PX1rTT/tddeo6SkhM8//1x32X7jxo0sXLgQgNTUVNRqNc7OzvoKWQghhKjXVIqiKPrY0bXW9WfPnkVRFObPn8/JkycpLCwkICCAUaNGER4ejkqlAmDSpEn07t2bOXPmkJycjEql4vnnnyc0NLTK/dzOr7aG8iuvrpG8G4bk3XAk94bRUPJe1Zm83oq8vkiRr/sk74YheTccyb1hNJS814nL9UIIIYTQLynyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJHSW5HXarW89tprREREMHHiRBISEirN3717N6NGjSIiIoL169dXax0hhBBC3JzeivzOnTspLS1l3bp1PPfccyxcuFA3r6ysjAULFrBixQrWrFnDunXrSE9Pr3IdIYQQQlTNTF87OnbsGD179gQgODiY2NhY3bwLFy7g4eFBkyZNAAgLC+Po0aNERUXddB0hhBBCVE1vRV6tVmNra6t7bWpqikajwczMDLVajZ2dnW6ejY0NarW6ynVuxtnZ7qbzamJ5UTMk74YheTccyb1hNPS86+1yva2tLQUFBbrXWq1WV6z/Pa+goAA7O7sq1xFCCCFE1fRW5ENDQ9m/fz8AUVFR+Pr66ub5+PiQkJBATk4OpaWlHD16lJCQkCrXEUIIIUTVVIqiKPrYkVarZe7cuZw9exZFUZg/fz4nT56ksLCQiIgIdu/ezWeffYaiKIwaNYoJEybccB0fHx99hCuEEELUe3or8kIIIYTQL+kMRwghhDBSUuSFEEIII9Xgmqpfu89/5swZLCwsmDdvHp6enoYOy6iNGDFC94iku7s7TzzxBLNnz0alUtGmTRtef/11TEzk92ZNiY6O5v3332fNmjUkJCTcMNfr16/n+++/x8zMjGnTptG3b19Dh13v/TPvcXFxPPHEE7Rq1QqA8ePHM2TIEMl7DSsrK+Pll18mKSmJ0tJSpk2bRuvWreWY/yelgfn999+Vl156SVEURYmMjFSeeOIJA0dk3IqLi5X777+/0rSpU6cqhw8fVhRFUf773/8q27dvN0BkxmnZsmXKsGHDlDFjxiiKcuNcp6WlKcOGDVNKSkqUvLw83f/Fnft33tevX6989dVXlZaRvNe8jRs3KvPmzVMURVGysrKU3r17yzH/Lw3u9KmqnvdEzTt9+jRFRUVMnjyZSZMmERUVRVxcHJ06dQKgV69e/PnnnwaO0nh4eHiwaNEi3esb5TomJoaQkBAsLCyws7PDw8OD06dPGypko/DvvMfGxrJ3714mTJjAyy+/jFqtlrzXgsGDB/PMM8/oXpuamsox/y8NrsjfrBc9UTusrKyYMmUKX331FW+88QbPP/88iqKgUqmAit4N8/PzDRyl8Rg0aFClDqNulOub9TAp7ty/8x4UFMSLL77I2rVradmyJZ999pnkvRbY2Nhga2uLWq3m6aef5tlnn5Vj/l8aXJGXXvT0y8vLi/vuuw+VSoWXlxf29vZkZmbq5hcUFNC4cWMDRmjc/tnW4Vqub9bDpKg5AwcOJCAgQPf/kydPSt5rSUpKCpMmTeL+++9n+PDhcsz/S4Mr8tKLnn5t3LhRN3pgamoqarWa7t2789dffwGwf/9+wsPDDRmiUfPz87su10FBQRw7doySkhLy8/O5cOGCfA5q2JQpU4iJiQHg0KFD+Pv7S95rQUZGBpMnT+aFF15g9OjRgBzz/9bgOsORXvT0q7S0lDlz5pCcnIxKpeL555/HwcGB//73v5SVleHt7c28efMwNTU1dKhGIzExkVmzZrF+/XouXrx4w1yvX7+edevWoSgKU6dOZdCgQYYOu977Z97j4uJ46623MDc3p2nTprz11lvY2tpK3mvYvHnz2Lp1K97e3rppr7zyCvPmzZNj/n8aXJEXQgghGooGd7leCCGEaCikyAshhBBGSoq8EEIIYaSkyAshhBBGSoq8EEIIYaSkyAthQLNnz6Zt27Y3/ffjjz/e0Taff/75ai07ceJEPvroo9veR23btm0b6enpt73e7bx3IRoCeYROCAPKz8+nuLgYgKNHj/Lss89y4MAB3Xw7OzusrKxue5vX1r2VnJwczM3NsbGxua191KakpCT69evH9u3bb3uEyNt570I0BNKfqxAGZGdnpytITZo0AcDZ2fmut1ld9vb2d7Wv2nA35x1S3IWoTC7XC1HHtW3blo8//pguXbrwyCOPAPDDDz9w7733EhAQQOfOnXn99dd1Ay3985L1okWLmDlzJm+++SZhYWH07duXpUuX6rb9z8v1s2fPZt68ecyaNYvg4GAGDRpU6XZBcXExr7zyCmFhYfTs2ZMNGzbg5+dHYmLiDeNeu3Yt/fv3JzAwkOHDh7Nnzx7dvKtXr/Lkk08SHBxMnz59eP/99yktLQWgf//+ANxzzz03vF2RkpLCf/7zH0JDQ+nUqRNz5szR9Uv+z/fer1+/G94CuWbdunX079+fkJAQxo8fr+uGVghjIkVeiHpg165dfPvtt7zyyiscPXqUN954g5kzZ/L777/zxhtv8OOPP7J9+/Ybrrtjxw5MTU3ZsGEDERERfPjhh5w/f/6Gy37//fe0b9+eH3/8kR49ejB37lxycnKAii5Ejx07xpdffslHH33El19+SXl5+Q23c/LkSRYsWMCcOXPYtm0bQ4YM4dlnnyUvLw9FUZg+fTpNmjThhx9+4P3332fv3r18+OGHAGzYsAGoKMJDhgy5bttvvvkmZmZm/PDDD6xYsYLIyEi++OKL65bbuHEjBw4c4MCBA+zYsQM3NzcmT54MwO7du/nkk0+YM2cOP/30E7169eLhhx8mLS2t6j+EEPWMFHkh6oGIiAi8vb1p06YNVlZWvP3229xzzz24ubkxePBg/Pz8blq47ezsmD17Nt7e3jzxxBPY29sTGxt7w2V9fX157LHH8Pb2ZubMmZSUlHDu3DkKCgr4+eefefXVVwkJCSE8PJxXX331pvEmJSUB4ObmhpubG1OnTuWzzz7D3Nycw4cPk5iYyLx58/Dx8SE8PJzXXnuNb775Bo1Gg6OjIwAODg43bI+QlJSEnZ0dbm5uBAQEsHjxYkaMGHHdco6Ojjg7O+Ps7MzHH3+Mi4sLzz33HABffvkljz/+OAMGDKBVq1ZMmzaNgIAA3Q8MIYyF3JMXoh5wc3PT/T8gIAArKys+/fRTzp8/z5kzZ0hISKBLly43XfefAwDZ2NhQVlZ2w2Vbtmyp+7+trS0AGo2G+Ph4ysrKCAwM1M0PCQm5abw9evQgLCyMESNG4OvrS79+/Rg9ejTW1tZcuHCBvLy8SqMPKopCWVkZycnJlYYKvZGnn36amTNnsmvXLnr06ME999xzwzP+a1avXs2ff/7Jzz//rBtW+sKFC3z44Yd88sknuuVKS0tp1qxZlfsWor6RIi9EPWBpaan7/x9//MGTTz7JiBEj6NmzJ9OnT+eNN9646brm5ubV3s+NllUURVcc/9korqoGctbW1nz99dccO3aMPXv2sG3bNr755hvWrl2LRqPB09OzUtuAa5o1a3bLS+YDBgxg37597Ny5k/379zNnzhwOHDigG9L4n44fP857773H559/XqmAl5eX89JLL9GjR49Kyzdq1KjKfQtR38jleiHqmQ0bNjBy5EjeeustxowZg4+PD5cvX67VfXp4eGBubk5cXJxu2s0u+QNERkby+eefEx4ezgsvvMDWrVtp2rQp+/fvx8vLi6tXr2Jvb4+npyeenp6kp6fzwQcfoCgKKpWqylg++ugjrl69ytixY1m8eDHz5s1jy5Yt1y2XmZnJM888w5QpU+jZs2eleddiuLZ/T09PVqxYwd9//32bmRGibpMiL0Q9Y29vT2RkJKdPn+bcuXPMnj2b9PR0Xev02mBjY8MDDzzAggULiIqKIioqirfffhvghkXZysqKzz//nO+//57ExER2795NSkoKAQEB9OjRg5YtW/L8889z+vRpIiMjefXVVzExMcHS0lJ3Nn369Gldq/l/io+P58033+TkyZPEx8ezfft2/P39Ky1TXl7Os88+S6tWrZg4cSLp6em6f6WlpTz66KOsWbOGn376icuXL7N48WJ++OGHSuOSC2EMpMgLUc/MmDEDFxcXxo0bx6OPPoq5uTkTJkzg5MmTtbrfl156iXbt2vHoo4/y1FNPMXz4cODGl/jbt2/PggULWLVqFffeey8LFizgpZdeolu3bpiamvL5559jamrKuHHjeOKJJwgPD2fevHlARYO7Bx54gOeee46NGzdet+25c+fi6urKI488wgMPPEB5eTkffPBBpWVSUlL4+++/+fvvv+nWrRs9evTQ/YuMjGTIkCE899xzLF68mKFDh7Jjxw4+++wz2rdvXwuZE8JwpMc7IUS17Ny5k65du+p6x4uJieHBBx8kMjLytu77CyH0RxreCSGqZfHixezevZupU6dSUFDAe++9R79+/aTAC1GHyZm8EKJazp8/z1tvvUVMTAwWFhb069ePl19+WbqSFaIOkyIvhBBCGClpeCeEEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJGSIi+EEEIYKSnyQgghhJH6P0HwSN8OznEoAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 576x396 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.style.use('seaborn')\n",
    "plt.plot(train_sizes, train_scores_mean, label = 'Training error')\n",
    "plt.plot(train_sizes, validation_scores_mean, label = 'Validation error')\n",
    "plt.ylabel('MSE', fontsize = 14)\n",
    "plt.xlabel('Training set size', fontsize = 14)\n",
    "plt.title('Learning curves for MLP model', fontsize = 18, y = 1.03)\n",
    "plt.legend()\n",
    "plt.ylim(0,2)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1b8966ed-491a-4a77-b5fe-ab64e86e7b48",
   "metadata": {},
   "source": [
    "expect at the smallest and largest training set sizes, the model is making the same amount of error while using both training and validation set suggesting that model is not improving at all as the size changes. Since the error is high, the model has a high bias and because the error scores for both training and validation are close to each other low variance meaning the model is under fitting the data. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "id": "a5d05025-fd0e-4f14-bc84-11d74fdd3255",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['MLP_model2.pkl']"
      ]
     },
     "execution_count": 128,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "joblib.dump(cv.best_estimator_, 'MLP_model2.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fb911a52-02a2-4af7-82fe-09044adc0170",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### **RF**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "id": "9d124ead-a490-479f-a43f-a9967a589e80",
   "metadata": {},
   "outputs": [],
   "source": [
    "import joblib\n",
    "import pandas as pd\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore', category=FutureWarning)\n",
    "warnings.filterwarnings('ignore', category=DeprecationWarning)\n",
    "\n",
    "tr_features = pd.read_csv('train_features2.csv')\n",
    "tr_labels = pd.read_csv('train_labels2.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "757bfa71-220f-4c7d-8982-21dccc6cfec9",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Hyperparameter tuning: n-estimators, max-depth"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "id": "8cd879a2-71c6-4128-a523-34bfe19755f9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BEST PARAMS: {'max_depth': 32, 'n_estimators': 250}\n",
      "\n",
      "0.931 (+/-0.106) for {'max_depth': 2, 'n_estimators': 5}\n",
      "0.967 (+/-0.015) for {'max_depth': 2, 'n_estimators': 50}\n",
      "0.967 (+/-0.027) for {'max_depth': 2, 'n_estimators': 250}\n",
      "0.949 (+/-0.081) for {'max_depth': 4, 'n_estimators': 5}\n",
      "0.971 (+/-0.037) for {'max_depth': 4, 'n_estimators': 50}\n",
      "0.971 (+/-0.018) for {'max_depth': 4, 'n_estimators': 250}\n",
      "0.96 (+/-0.027) for {'max_depth': 8, 'n_estimators': 5}\n",
      "0.971 (+/-0.037) for {'max_depth': 8, 'n_estimators': 50}\n",
      "0.978 (+/-0.042) for {'max_depth': 8, 'n_estimators': 250}\n",
      "0.971 (+/-0.037) for {'max_depth': 16, 'n_estimators': 5}\n",
      "0.971 (+/-0.018) for {'max_depth': 16, 'n_estimators': 50}\n",
      "0.975 (+/-0.044) for {'max_depth': 16, 'n_estimators': 250}\n",
      "0.978 (+/-0.036) for {'max_depth': 32, 'n_estimators': 5}\n",
      "0.978 (+/-0.042) for {'max_depth': 32, 'n_estimators': 50}\n",
      "0.982 (+/-0.033) for {'max_depth': 32, 'n_estimators': 250}\n",
      "0.967 (+/-0.027) for {'max_depth': None, 'n_estimators': 5}\n",
      "0.971 (+/-0.037) for {'max_depth': None, 'n_estimators': 50}\n",
      "0.982 (+/-0.033) for {'max_depth': None, 'n_estimators': 250}\n"
     ]
    }
   ],
   "source": [
    "rf = RandomForestClassifier()\n",
    "parameters = {\n",
    "    'n_estimators': [5, 50, 250],\n",
    "    'max_depth': [2, 4, 8, 16, 32, None]\n",
    "}\n",
    "\n",
    "cv = GridSearchCV(rf, parameters, cv=5)\n",
    "cv.fit(tr_features, tr_labels.values.ravel())\n",
    "\n",
    "print_results(cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b239fdf2-57a6-4282-b5ae-fe4bf100e548",
   "metadata": {},
   "source": [
    "both n_estimator parameter and max_depth are affecting the performance of model with n-estimators having a greater effect. When max_depth = 32 and there are 250 estimators, the model achieves its highest performance of 98.2%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "id": "63f7a947-63f3-4049-9512-7f1ea559cebf",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(max_depth=32, n_estimators=250)"
      ]
     },
     "execution_count": 151,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "88552441-2d7e-4604-8a19-26f5dfd7658f",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### learning curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "id": "7d0230ae-9319-4bbf-b919-9e7584aeb3ad",
   "metadata": {},
   "outputs": [],
   "source": [
    "#citation 3\n",
    "from sklearn.model_selection import learning_curve\n",
    "# the training data is 80% of the whole data --> 275\n",
    "# we are doing 5-fold cross validation with 80% training set and 20% validation set:\n",
    "# 275 training example : 80% training (220) 20% validation (55) in cross validation\n",
    "train_sizes = [5, 50, 100, 150, 200, 220]\n",
    "train_sizes, train_scores, validation_scores = learning_curve(\n",
    "estimator = RandomForestClassifier(max_depth=8, n_estimators=250),\n",
    "X = tr_features,\n",
    "y = tr_labels.values.ravel(), train_sizes = train_sizes, cv = 5,\n",
    "scoring = 'neg_mean_squared_error')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "id": "3c3e30ae-86e5-4bd6-b790-65921b7c451f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean training scores\n",
      "\n",
      " 5     -0.0\n",
      "50    -0.0\n",
      "100   -0.0\n",
      "150   -0.0\n",
      "200   -0.0\n",
      "220   -0.0\n",
      "dtype: float64\n",
      "\n",
      " --------------------\n",
      "\n",
      "Mean validation scores\n",
      "\n",
      " 5      0.276364\n",
      "50     0.072727\n",
      "100    0.061818\n",
      "150    0.050909\n",
      "200    0.032727\n",
      "220    0.047273\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "train_scores_mean = -train_scores.mean(axis = 1)\n",
    "validation_scores_mean = -validation_scores.mean(axis = 1)\n",
    "print('Mean training scores\\n\\n', pd.Series(train_scores_mean, index = train_sizes))\n",
    "print('\\n', '-' * 20) # separator\n",
    "print('\\nMean validation scores\\n\\n',pd.Series(validation_scores_mean, index = train_sizes))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7d6ed646-25d7-4683-ae6e-b6d7f1ddbfff",
   "metadata": {},
   "source": [
    "like previous model The validation error is high for the smaller set as it should and as the model starts making better generalization as the training size increases, the validation error gradually decreases however when the sample size exceeds 200 the validation starts increasing than decreasing. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "id": "af0c8b83-570d-4c55-97e0-b0e5b3e1769b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 0.5)"
      ]
     },
     "execution_count": 155,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x396 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.style.use('seaborn')\n",
    "plt.plot(train_sizes, train_scores_mean, label = 'Training error')\n",
    "plt.plot(train_sizes, validation_scores_mean, label = 'Validation error')\n",
    "plt.ylabel('MSE', fontsize = 14)\n",
    "plt.xlabel('Training set size', fontsize = 14)\n",
    "plt.title('Learning curves for RF model', fontsize = 18, y = 1.03)\n",
    "plt.legend()\n",
    "plt.ylim(0,0.5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2526336a-7d83-4aeb-a802-8221b1275708",
   "metadata": {},
   "source": [
    "As discussed previously, the RF model overfits our data the slight increase increase in bias and variance does not change this fact however there is a possibility that gather more actual data and increasing the overall size of our dataset might improve this model's performance. But for the current situation, this model is not suitable to solve our problem. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "id": "38fc028e-4af9-4bf9-96d3-f84550834dd9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['RF_model2.pkl']"
      ]
     },
     "execution_count": 156,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "joblib.dump(cv.best_estimator_, 'RF_model2.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "07f6df16-e58f-4846-bd75-d73535598261",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### **Boosting**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8313f1f5-0fb8-4afa-9997-d2b6e67e169b",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Hyperparameter Tunning: n-estimators, max_depth, & learning_rate Tuning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "id": "2f2ae464-278b-49cc-9c05-48506163111e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BEST PARAMS: {'learning_rate': 1, 'max_depth': 3, 'n_estimators': 50}\n",
      "\n",
      "0.436 (+/-0.0) for {'learning_rate': 0.01, 'max_depth': 1, 'n_estimators': 5}\n",
      "0.938 (+/-0.037) for {'learning_rate': 0.01, 'max_depth': 1, 'n_estimators': 50}\n",
      "0.956 (+/-0.037) for {'learning_rate': 0.01, 'max_depth': 1, 'n_estimators': 250}\n",
      "0.967 (+/-0.027) for {'learning_rate': 0.01, 'max_depth': 1, 'n_estimators': 500}\n",
      "0.436 (+/-0.0) for {'learning_rate': 0.01, 'max_depth': 3, 'n_estimators': 5}\n",
      "0.975 (+/-0.018) for {'learning_rate': 0.01, 'max_depth': 3, 'n_estimators': 50}\n",
      "0.975 (+/-0.018) for {'learning_rate': 0.01, 'max_depth': 3, 'n_estimators': 250}\n",
      "0.971 (+/-0.018) for {'learning_rate': 0.01, 'max_depth': 3, 'n_estimators': 500}\n",
      "0.436 (+/-0.0) for {'learning_rate': 0.01, 'max_depth': 5, 'n_estimators': 5}\n",
      "0.975 (+/-0.037) for {'learning_rate': 0.01, 'max_depth': 5, 'n_estimators': 50}\n",
      "0.971 (+/-0.037) for {'learning_rate': 0.01, 'max_depth': 5, 'n_estimators': 250}\n",
      "0.971 (+/-0.037) for {'learning_rate': 0.01, 'max_depth': 5, 'n_estimators': 500}\n",
      "0.436 (+/-0.0) for {'learning_rate': 0.01, 'max_depth': 7, 'n_estimators': 5}\n",
      "0.975 (+/-0.037) for {'learning_rate': 0.01, 'max_depth': 7, 'n_estimators': 50}\n",
      "0.971 (+/-0.037) for {'learning_rate': 0.01, 'max_depth': 7, 'n_estimators': 250}\n",
      "0.975 (+/-0.037) for {'learning_rate': 0.01, 'max_depth': 7, 'n_estimators': 500}\n",
      "0.436 (+/-0.0) for {'learning_rate': 0.01, 'max_depth': 9, 'n_estimators': 5}\n",
      "0.975 (+/-0.037) for {'learning_rate': 0.01, 'max_depth': 9, 'n_estimators': 50}\n",
      "0.971 (+/-0.037) for {'learning_rate': 0.01, 'max_depth': 9, 'n_estimators': 250}\n",
      "0.975 (+/-0.037) for {'learning_rate': 0.01, 'max_depth': 9, 'n_estimators': 500}\n",
      "0.942 (+/-0.042) for {'learning_rate': 0.1, 'max_depth': 1, 'n_estimators': 5}\n",
      "0.967 (+/-0.027) for {'learning_rate': 0.1, 'max_depth': 1, 'n_estimators': 50}\n",
      "0.978 (+/-0.042) for {'learning_rate': 0.1, 'max_depth': 1, 'n_estimators': 250}\n",
      "0.975 (+/-0.054) for {'learning_rate': 0.1, 'max_depth': 1, 'n_estimators': 500}\n",
      "0.971 (+/-0.029) for {'learning_rate': 0.1, 'max_depth': 3, 'n_estimators': 5}\n",
      "0.971 (+/-0.018) for {'learning_rate': 0.1, 'max_depth': 3, 'n_estimators': 50}\n",
      "0.978 (+/-0.027) for {'learning_rate': 0.1, 'max_depth': 3, 'n_estimators': 250}\n",
      "0.975 (+/-0.037) for {'learning_rate': 0.1, 'max_depth': 3, 'n_estimators': 500}\n",
      "0.975 (+/-0.037) for {'learning_rate': 0.1, 'max_depth': 5, 'n_estimators': 5}\n",
      "0.971 (+/-0.037) for {'learning_rate': 0.1, 'max_depth': 5, 'n_estimators': 50}\n",
      "0.971 (+/-0.037) for {'learning_rate': 0.1, 'max_depth': 5, 'n_estimators': 250}\n",
      "0.964 (+/-0.04) for {'learning_rate': 0.1, 'max_depth': 5, 'n_estimators': 500}\n",
      "0.975 (+/-0.037) for {'learning_rate': 0.1, 'max_depth': 7, 'n_estimators': 5}\n",
      "0.971 (+/-0.037) for {'learning_rate': 0.1, 'max_depth': 7, 'n_estimators': 50}\n",
      "0.971 (+/-0.037) for {'learning_rate': 0.1, 'max_depth': 7, 'n_estimators': 250}\n",
      "0.964 (+/-0.04) for {'learning_rate': 0.1, 'max_depth': 7, 'n_estimators': 500}\n",
      "0.975 (+/-0.037) for {'learning_rate': 0.1, 'max_depth': 9, 'n_estimators': 5}\n",
      "0.975 (+/-0.037) for {'learning_rate': 0.1, 'max_depth': 9, 'n_estimators': 50}\n",
      "0.971 (+/-0.037) for {'learning_rate': 0.1, 'max_depth': 9, 'n_estimators': 250}\n",
      "0.964 (+/-0.04) for {'learning_rate': 0.1, 'max_depth': 9, 'n_estimators': 500}\n",
      "0.971 (+/-0.029) for {'learning_rate': 1, 'max_depth': 1, 'n_estimators': 5}\n",
      "0.978 (+/-0.042) for {'learning_rate': 1, 'max_depth': 1, 'n_estimators': 50}\n",
      "0.978 (+/-0.042) for {'learning_rate': 1, 'max_depth': 1, 'n_estimators': 250}\n",
      "0.978 (+/-0.042) for {'learning_rate': 1, 'max_depth': 1, 'n_estimators': 500}\n",
      "0.975 (+/-0.018) for {'learning_rate': 1, 'max_depth': 3, 'n_estimators': 5}\n",
      "0.982 (+/-0.023) for {'learning_rate': 1, 'max_depth': 3, 'n_estimators': 50}\n",
      "0.978 (+/-0.027) for {'learning_rate': 1, 'max_depth': 3, 'n_estimators': 250}\n",
      "0.975 (+/-0.018) for {'learning_rate': 1, 'max_depth': 3, 'n_estimators': 500}\n",
      "0.96 (+/-0.015) for {'learning_rate': 1, 'max_depth': 5, 'n_estimators': 5}\n",
      "0.964 (+/-0.023) for {'learning_rate': 1, 'max_depth': 5, 'n_estimators': 50}\n",
      "0.953 (+/-0.029) for {'learning_rate': 1, 'max_depth': 5, 'n_estimators': 250}\n",
      "0.949 (+/-0.048) for {'learning_rate': 1, 'max_depth': 5, 'n_estimators': 500}\n",
      "0.96 (+/-0.015) for {'learning_rate': 1, 'max_depth': 7, 'n_estimators': 5}\n",
      "0.964 (+/-0.023) for {'learning_rate': 1, 'max_depth': 7, 'n_estimators': 50}\n",
      "0.96 (+/-0.015) for {'learning_rate': 1, 'max_depth': 7, 'n_estimators': 250}\n",
      "0.964 (+/-0.033) for {'learning_rate': 1, 'max_depth': 7, 'n_estimators': 500}\n",
      "0.96 (+/-0.015) for {'learning_rate': 1, 'max_depth': 9, 'n_estimators': 5}\n",
      "0.96 (+/-0.015) for {'learning_rate': 1, 'max_depth': 9, 'n_estimators': 50}\n",
      "0.96 (+/-0.015) for {'learning_rate': 1, 'max_depth': 9, 'n_estimators': 250}\n",
      "0.96 (+/-0.027) for {'learning_rate': 1, 'max_depth': 9, 'n_estimators': 500}\n",
      "0.502 (+/-0.396) for {'learning_rate': 10, 'max_depth': 1, 'n_estimators': 5}\n",
      "0.502 (+/-0.396) for {'learning_rate': 10, 'max_depth': 1, 'n_estimators': 50}\n",
      "0.502 (+/-0.396) for {'learning_rate': 10, 'max_depth': 1, 'n_estimators': 250}\n",
      "0.502 (+/-0.396) for {'learning_rate': 10, 'max_depth': 1, 'n_estimators': 500}\n",
      "0.542 (+/-0.705) for {'learning_rate': 10, 'max_depth': 3, 'n_estimators': 5}\n",
      "0.644 (+/-0.725) for {'learning_rate': 10, 'max_depth': 3, 'n_estimators': 50}\n",
      "0.647 (+/-0.731) for {'learning_rate': 10, 'max_depth': 3, 'n_estimators': 250}\n",
      "0.647 (+/-0.731) for {'learning_rate': 10, 'max_depth': 3, 'n_estimators': 500}\n",
      "0.956 (+/-0.059) for {'learning_rate': 10, 'max_depth': 5, 'n_estimators': 5}\n",
      "0.956 (+/-0.044) for {'learning_rate': 10, 'max_depth': 5, 'n_estimators': 50}\n",
      "0.956 (+/-0.049) for {'learning_rate': 10, 'max_depth': 5, 'n_estimators': 250}\n",
      "0.945 (+/-0.033) for {'learning_rate': 10, 'max_depth': 5, 'n_estimators': 500}\n",
      "0.96 (+/-0.042) for {'learning_rate': 10, 'max_depth': 7, 'n_estimators': 5}\n",
      "0.953 (+/-0.049) for {'learning_rate': 10, 'max_depth': 7, 'n_estimators': 50}\n",
      "0.953 (+/-0.049) for {'learning_rate': 10, 'max_depth': 7, 'n_estimators': 250}\n",
      "0.949 (+/-0.053) for {'learning_rate': 10, 'max_depth': 7, 'n_estimators': 500}\n",
      "0.956 (+/-0.044) for {'learning_rate': 10, 'max_depth': 9, 'n_estimators': 5}\n",
      "0.953 (+/-0.054) for {'learning_rate': 10, 'max_depth': 9, 'n_estimators': 50}\n",
      "0.956 (+/-0.049) for {'learning_rate': 10, 'max_depth': 9, 'n_estimators': 250}\n",
      "0.953 (+/-0.049) for {'learning_rate': 10, 'max_depth': 9, 'n_estimators': 500}\n",
      "0.425 (+/-0.266) for {'learning_rate': 100, 'max_depth': 1, 'n_estimators': 5}\n",
      "0.425 (+/-0.266) for {'learning_rate': 100, 'max_depth': 1, 'n_estimators': 50}\n",
      "0.425 (+/-0.266) for {'learning_rate': 100, 'max_depth': 1, 'n_estimators': 250}\n",
      "0.425 (+/-0.266) for {'learning_rate': 100, 'max_depth': 1, 'n_estimators': 500}\n",
      "0.716 (+/-0.526) for {'learning_rate': 100, 'max_depth': 3, 'n_estimators': 5}\n",
      "0.829 (+/-0.332) for {'learning_rate': 100, 'max_depth': 3, 'n_estimators': 50}\n",
      "0.731 (+/-0.506) for {'learning_rate': 100, 'max_depth': 3, 'n_estimators': 250}\n",
      "0.716 (+/-0.526) for {'learning_rate': 100, 'max_depth': 3, 'n_estimators': 500}\n",
      "0.949 (+/-0.048) for {'learning_rate': 100, 'max_depth': 5, 'n_estimators': 5}\n",
      "0.956 (+/-0.029) for {'learning_rate': 100, 'max_depth': 5, 'n_estimators': 50}\n",
      "0.96 (+/-0.036) for {'learning_rate': 100, 'max_depth': 5, 'n_estimators': 250}\n",
      "0.953 (+/-0.037) for {'learning_rate': 100, 'max_depth': 5, 'n_estimators': 500}\n",
      "0.945 (+/-0.046) for {'learning_rate': 100, 'max_depth': 7, 'n_estimators': 5}\n",
      "0.953 (+/-0.037) for {'learning_rate': 100, 'max_depth': 7, 'n_estimators': 50}\n",
      "0.956 (+/-0.029) for {'learning_rate': 100, 'max_depth': 7, 'n_estimators': 250}\n",
      "0.956 (+/-0.029) for {'learning_rate': 100, 'max_depth': 7, 'n_estimators': 500}\n",
      "0.956 (+/-0.044) for {'learning_rate': 100, 'max_depth': 9, 'n_estimators': 5}\n",
      "0.949 (+/-0.048) for {'learning_rate': 100, 'max_depth': 9, 'n_estimators': 50}\n",
      "0.956 (+/-0.029) for {'learning_rate': 100, 'max_depth': 9, 'n_estimators': 250}\n",
      "0.953 (+/-0.029) for {'learning_rate': 100, 'max_depth': 9, 'n_estimators': 500}\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "\n",
    "gb = GradientBoostingClassifier()\n",
    "parameters = {\n",
    "    'n_estimators': [5, 50, 250, 500],\n",
    "    'max_depth': [1, 3, 5, 7, 9],\n",
    "    'learning_rate': [0.01, 0.1, 1, 10, 100]\n",
    "}\n",
    "\n",
    "cv = GridSearchCV(gb, parameters, cv=5)\n",
    "cv.fit(tr_features, tr_labels.values.ravel())\n",
    "\n",
    "print_results(cv)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2c24d2fb-f663-4638-a46e-81ae89d0f448",
   "metadata": {},
   "source": [
    "overall the model has the best performance (98.2%) with learning rate of 1, max_depth of 3 and 50 estimators which is slightly lower than the GB model in the previous split ratio. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 158,
   "id": "96012b88-4348-4ab7-88c1-f10ebf526800",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "GradientBoostingClassifier(learning_rate=1, n_estimators=50)"
      ]
     },
     "execution_count": 158,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4a7f2e21-3de8-4d42-b8c0-afb8a895e59f",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Learning Curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "id": "3fb42260-8f9a-498d-b911-6cca601f837c",
   "metadata": {},
   "outputs": [],
   "source": [
    "#citation 3\n",
    "from sklearn.model_selection import learning_curve\n",
    "# the training data is 80% of the whole data --> 275\n",
    "# we are doing 5-fold cross validation with 80% training set and 20% validation set:\n",
    "# 275 training example : 80% training (220) 20% validation (55) in cross validation\n",
    "train_sizes = [5, 50, 100, 150, 200, 220]\n",
    "train_sizes, train_scores, validation_scores = learning_curve(\n",
    "estimator = GradientBoostingClassifier(learning_rate=1, n_estimators=50),\n",
    "X = tr_features,\n",
    "y = tr_labels.values.ravel(), train_sizes = train_sizes, cv = 5,\n",
    "scoring = 'neg_mean_squared_error')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "id": "0373d0ee-917b-4773-9567-423102f4fac2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean training scores\n",
      "\n",
      " 5     -0.0\n",
      "50    -0.0\n",
      "100   -0.0\n",
      "150   -0.0\n",
      "200   -0.0\n",
      "220   -0.0\n",
      "dtype: float64\n",
      "\n",
      " --------------------\n",
      "\n",
      "Mean validation scores\n",
      "\n",
      " 5      0.363636\n",
      "50     0.152727\n",
      "100    0.047273\n",
      "150    0.061818\n",
      "200    0.036364\n",
      "220    0.029091\n",
      "dtype: float64\n"
     ]
    }
   ],
   "source": [
    "train_scores_mean = -train_scores.mean(axis = 1)\n",
    "validation_scores_mean = -validation_scores.mean(axis = 1)\n",
    "print('Mean training scores\\n\\n', pd.Series(train_scores_mean, index = train_sizes))\n",
    "print('\\n', '-' * 20) # separator\n",
    "print('\\nMean validation scores\\n\\n',pd.Series(validation_scores_mean, index = train_sizes))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "id": "c89f6623-ecc1-43d6-a25d-f90bc745a94b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 0.6)"
      ]
     },
     "execution_count": 161,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x396 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.style.use('seaborn')\n",
    "plt.plot(train_sizes, train_scores_mean, label = 'Training error')\n",
    "plt.plot(train_sizes, validation_scores_mean, label = 'Validation error')\n",
    "plt.ylabel('MSE', fontsize = 14)\n",
    "plt.xlabel('Training set size', fontsize = 14)\n",
    "plt.title('Learning curves for GB model', fontsize = 18, y = 1.03)\n",
    "plt.legend()\n",
    "plt.ylim(0,0.6)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7e87aaba-73ab-4378-8be7-9af9141f47bc",
   "metadata": {},
   "source": [
    "Compare to the GB model from the previous split, the model has slightly higher variance suggesting that even when increasing the data point in the training set, GB model still overfits the data and not suitable to solve our problem"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "id": "bffe04f2-8320-4dc8-a7aa-8d422b163cf1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['GB_model2.pkl']"
      ]
     },
     "execution_count": 162,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "joblib.dump(cv.best_estimator_, 'GB_model2.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c54aaf11-b882-471b-bca9-bfb03b7fe1d8",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "### **Model comparison**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ce27ef6b-1b91-4f3c-91da-666477ca76d5",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Read in Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "id": "18cbe137-3dfd-43fa-aeb5-a4d8f611767e",
   "metadata": {},
   "outputs": [],
   "source": [
    "import joblib\n",
    "import pandas as pd\n",
    "from sklearn.metrics import accuracy_score, precision_score, recall_score\n",
    "from time import time\n",
    "\n",
    "val_features = pd.read_csv('val_features2.csv')\n",
    "val_labels = pd.read_csv('val_labels2.csv')\n",
    "\n",
    "te_features = pd.read_csv('test_features2.csv')\n",
    "te_labels = pd.read_csv('test_labels2.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d72e3d54-4446-453f-86ff-01294994c597",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Read in Models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "id": "b9e5a44e-4bc2-4480-a758-f4fb7db253d2",
   "metadata": {},
   "outputs": [],
   "source": [
    "models = {}\n",
    "\n",
    "for mdl in ['LR', 'SVM', 'MLP', 'RF', 'GB']:\n",
    "    models[mdl] = joblib.load('{}_model2.pkl'.format(mdl))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "id": "2d38d03b-e4e5-4c6b-a344-858e63e687a4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'LR': LogisticRegression(C=1, max_iter=4500),\n",
       " 'SVM': SVC(C=0.1, kernel='linear'),\n",
       " 'MLP': MLPClassifier(activation='logistic', max_iter=2500),\n",
       " 'RF': RandomForestClassifier(max_depth=32, n_estimators=250),\n",
       " 'GB': GradientBoostingClassifier(learning_rate=1, n_estimators=50)}"
      ]
     },
     "execution_count": 165,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "models"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d0e12929-e6a2-42aa-aaa0-86d1edd34170",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### Evaluate models on the validation set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "id": "df4f8a60-8467-4b35-a17e-7ed466fa879f",
   "metadata": {},
   "outputs": [],
   "source": [
    "def evaluate_model(name, model, features, labels):\n",
    "    start = time()\n",
    "    pred = model.predict(features)\n",
    "    end = time()\n",
    "    latency = round((end - start)*1000, 1)\n",
    "    accuracy = round(accuracy_score(labels, pred), 3)\n",
    "    precision = round(precision_score(labels, pred, average='macro'), 3)\n",
    "    recall = round(recall_score(labels, pred, average='macro'), 3)\n",
    "    print('{} --> Accuracy= {} | Precision= {} | Recall= {} | Latency= {}ms'.format(name, accuracy, precision, recall, latency ))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "id": "9ffb45a9-8391-4658-b1bf-a14b821b79aa",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR --> Accuracy= 1.0 | Precision= 1.0 | Recall= 1.0 | Latency= 1.0ms\n",
      "SVM --> Accuracy= 1.0 | Precision= 1.0 | Recall= 1.0 | Latency= 2.0ms\n",
      "MLP --> Accuracy= 0.971 | Precision= 0.972 | Recall= 0.982 | Latency= 1.5ms\n",
      "RF --> Accuracy= 0.971 | Precision= 0.972 | Recall= 0.982 | Latency= 21.0ms\n",
      "GB --> Accuracy= 0.971 | Precision= 0.972 | Recall= 0.982 | Latency= 2.0ms\n"
     ]
    }
   ],
   "source": [
    "for name, mdl in models.items():\n",
    "    evaluate_model(name, mdl, val_features, val_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fb651d90-b623-4528-8508-bcc1dc9a478c",
   "metadata": {},
   "source": [
    "We can see that with the new split ratio, the overall accuracy score increased for all algorithms. The precision and recall increased for LR, SVM, and MLP while it decreased slightly for RF and GB. Based on this result again LR and SVM are both suitable for solving this problem with 100% accuracy, precision and recall scores. The only difference between the two model is in their latency where LR is twice as fast as SVM makeing it a better choice. \n",
    "\n",
    "evaluation result from 60:20:20\n",
    "\n",
    "LR --> Accuracy= 0.986 | Precision= 0.987 | Recall= 0.99 | Latency= 1.0ms\n",
    "\n",
    "SVM --> Accuracy= 0.986 | Precision= 0.99 | Recall= 0.972 | Latency= 0.8ms\n",
    "\n",
    "MLP --> Accuracy= 0.681 | Precision= 0.454 | Recall= 0.56 | Latency= 1.0ms\n",
    "\n",
    "RF --> Accuracy= 0.986 | Precision= 0.974 | Recall= 0.99 | Latency= 21.0ms\n",
    "\n",
    "GB --> Accuracy= 0.986 | Precision= 0.974 | Recall= 0.99 | Latency= 1.0ms\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "63a18f18-0b72-4dff-992d-2d60b57f5097",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "#### evaluate models on test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "id": "6b399d7c-11f0-407b-a0f3-198b0c188a0c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR --> Accuracy= 1.0 | Precision= 1.0 | Recall= 1.0 | Latency= 2.0ms\n",
      "SVM --> Accuracy= 0.971 | Precision= 0.976 | Recall= 0.97 | Latency= 1.0ms\n",
      "MLP --> Accuracy= 0.941 | Precision= 0.944 | Recall= 0.944 | Latency= 1.0ms\n",
      "RF --> Accuracy= 1.0 | Precision= 1.0 | Recall= 1.0 | Latency= 22.6ms\n",
      "GB --> Accuracy= 0.971 | Precision= 0.976 | Recall= 0.97 | Latency= 2.0ms\n"
     ]
    }
   ],
   "source": [
    "for name, mdl in models.items():\n",
    "    evaluate_model(name, mdl, te_features, te_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a5fe9bc2-87ea-4022-9ba4-67957c7551fc",
   "metadata": {},
   "source": [
    "by running the models on the test set we can see that LR and RF did better than rest of the models. However given the RF's lower scores when evaluating the model using validation set and it's tendency to overfit compared to LR during learning curve analysis, we can conclude that LR is best model when it comes to solving this problem with split ratio of 80:10:10"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b80fe198-1e63-4dcb-82d8-6097e275a230",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true,
    "tags": []
   },
   "source": [
    "## **Part V: Conclusion**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "99fa6644-446f-4d2f-893d-9c7985530bad",
   "metadata": {},
   "source": [
    "When splitting the data 60:20:20 we found that SVM model is the best model with accuracy of 98.6%, precision of 99%, recall of 98.7% and latency of 1.5 ms and when splitting the data 80:10:10 we found that LR model with accuracy, precision, and recall of 100% and latency of 2 ms was the best model to solve our problem. During our learning curve analysis we noticed that model performance drops  slightly for the higher splitting ratio compare to the lower one. We also noted that the latency almost doubles when we have higher ratio. \n",
    "\n",
    "Deciding on which split ratio is better than other depends on few factors. If the accuracy and precision of the model is of concern then split ratio of 80:10:10 is better choice even though the computation is slightly more expensive and takes longer for the model to make prediction. However, if the responsive of the model is concerned (for example the model need to make prediction in real time) the 60:20:20 with SVM model would be a better fit. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a9c9c895-0908-46f4-9416-f806e67a010f",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}