symptoms_data.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "47870620",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np \n",
    "import cvxopt\n",
    "from cvxopt import solvers\n",
    "from cvxopt import matrix\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from sklearn.metrics import roc_curve, auc\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn import svm as SVM\n",
    "import seaborn as sn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "d69a4f95",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_data = pd.read_csv(\"train.csv\", usecols = np.linspace(0,8,9,dtype=int)).to_numpy()\n",
    "train_label = pd.read_csv(\"train.csv\", usecols = [9]).to_numpy()\n",
    "\n",
    "test_data = pd.read_csv(\"test.csv\", usecols = np.linspace(0,8,9,dtype=int)).to_numpy()\n",
    "test_label = pd.read_csv(\"test.csv\", usecols = [9]).to_numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "efce0489",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(253440, 9)\n",
      "(63360, 9)\n"
     ]
    }
   ],
   "source": [
    "print(train_data.shape)\n",
    "print(test_data.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "a6779b42",
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in range(train_label.shape[0]):\n",
    "    if train_label[i][0] != 0:\n",
    "        train_label[i][0] = 1\n",
    "    else:\n",
    "        train_label[i][0] = -1\n",
    "\n",
    "for i in range(test_label.shape[0]):\n",
    "    if test_label[i][0] != 0:\n",
    "        test_label[i][0] = 1\n",
    "    else:\n",
    "        test_label[i][0] = -1"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9afe8db2",
   "metadata": {},
   "source": [
    "Since for $\\forall 1 \\leq i \\leq m $, where $m$ is the number of data points, our primal objective function with kernel and slack variables is\n",
    "\\begin{align*}\n",
    "\\min _{w, b, \\xi} & \\frac{1}{2} w^Tw+C \\sum_{i=1}^{m} \\xi_{i} \\\\\n",
    "\\text { s.t.}\\quad  & y_{i}\\left(w^{\\top} x_{i}+b\\right) \\geq 1-\\xi_{i} \\\\\n",
    "& \\xi_{i} \\geq 0\n",
    "\\end{align*}\n",
    "\n",
    "For an optimization problem of CVXOPT, we need the formation\n",
    "\\begin{align*}\n",
    "\\min _{x} & \\frac{1}{2} x^{T} P x+q^{T} x \\\\\n",
    "\\text { s.t.}\\quad  & G x \\leq h\n",
    "\\end{align*}\n",
    "\n",
    "So, denote $m$ as the number of data points and $n$ as the number of dimensions, we could construct the required matrices as\n",
    "\\begin{align*}\n",
    "P_1 &= \\left[\\begin{array}{cc}\n",
    "0 & 0 \\\\\n",
    "0 & I_n\n",
    "\\end{array}\\right]\\in \\mathbb{R}^{n+1}\\times \\mathbb{R}^{n+1}\\\\\n",
    "P &= \\left[\\begin{array}{cc}\n",
    "P_1 & 0_{(n+1) \\times m} \\\\\n",
    "0_{m\\times (n+1)} & 0_{m\\times m}\n",
    "\\end{array}\\right] \\in \\mathbb{R}^{m+n+1}\\times \\mathbb{R}^{m+n+1}\\\\\n",
    "q &= \\left[0,\\dots,0,C,\\dots,C\\right] \\in 1\\times \\mathbb{R}^{m+n+1}\\\\\n",
    "G_1^{(i)} &= -y^{(i)} x^{(i)}, \\text { where } x^{(i)} \\text { represents the } \\left[1 \\mid i^{th} \\text { row of the data matrix}\\right]\\\\\n",
    "G &= \\left[\\begin{array}{cc}\n",
    "G_1 & - I_{m} \\\\\n",
    "0_{m \\times (n+1)} & -I_{m}\n",
    "\\end{array}\\right] \\\\\n",
    "h &= [-1, \\ldots,-1,0, \\ldots, 0] \\text{ with m -1's and m 0's}\n",
    "\\end{align*}\n",
    "So, after we solve the quadratic program, we have $b = x_0$ and $w = x_{1:n}$"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f9aa0d20",
   "metadata": {},
   "source": [
    "###  Build primal form of soft margin SVM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "94c51238",
   "metadata": {},
   "outputs": [],
   "source": [
    "class binary_SVM:\n",
    "    \n",
    "    def __init__(self, C):\n",
    "        self.C = C\n",
    "        \n",
    "    def fit(self, train_data, train_label):\n",
    "        \"\"\"\n",
    "        Train SVM (primal form)\n",
    "\n",
    "        Argument:\n",
    "            train_data: N*D matrix, each row as a sample and each column as a feature\n",
    "            train_label: N*1 vector, each row as a label\n",
    "            C: tradeoff parameter (on slack variable side)\n",
    "\n",
    "        Return:\n",
    "            w: feature vector (column vector)\n",
    "            b: bias term\n",
    "        \"\"\"\n",
    "        m, n = train_data.shape\n",
    "        \n",
    "        P_1 = np.identity(n + 1)\n",
    "        P_1[0, 0] = 0\n",
    "        \n",
    "    \n",
    "        tmp_1 = np.zeros((n + 1, m))\n",
    "        tmp_2 = np.zeros((m, n + 1))\n",
    "        tmp_3 = np.zeros((m, m))\n",
    "    \n",
    "        P_u = np.concatenate((P_1,tmp_1),axis = 1)\n",
    "        P_d = np.concatenate((tmp_2,tmp_3),axis = 1)\n",
    "    \n",
    "        P = matrix(np.concatenate((P_u,P_d), axis = 0))\n",
    "        q = matrix(np.array([0.] * (n + 1) + [self.C] * m).reshape(-1,1))\n",
    "    \n",
    "        train_with_leading = np.pad(train_data, pad_width=((0, 0), (1, 0)), constant_values=1)\n",
    "        G_1 = -train_label * train_with_leading\n",
    "        tmp_4 = -np.identity(m)\n",
    "        tmp_5 = np.zeros((m, n + 1))\n",
    "        G_u = np.concatenate((G_1,tmp_4),axis = 1)\n",
    "        G_d = np.concatenate((tmp_5,tmp_4),axis = 1)\n",
    "        G = matrix(np.concatenate((G_u,G_d), axis = 0))\n",
    "    \n",
    "        h = matrix(np.array([-1.] * m + [0] * m).reshape(-1,1))\n",
    "    \n",
    "        solvers.options['show_progress'] = False\n",
    "        solution = solvers.qp(P, q, G, h)['x']\n",
    "    \n",
    "        self.w = np.array(solution[1:len(solution) - m])\n",
    "        self.b = solution[0]\n",
    "\n",
    "\n",
    "    def predict(self, test_data, test_label):\n",
    "        \"\"\"\n",
    "        Test linear SVM\n",
    "\n",
    "        Argument:\n",
    "            test_data: M*D matrix, each row as a sample and each column as a feature\n",
    "            test_label: M*1 vector, each row as a label\n",
    "            w: feature vector\n",
    "            b: bias term\n",
    "\n",
    "        Return:\n",
    "            test_accuracy: a float between [0, 1] representing the test accuracy\n",
    "        \"\"\"\n",
    "        m,n = test_data.shape\n",
    "        res = []\n",
    "        pred = np.dot(test_data, self.w) + self.b\n",
    "        for i in range(m):\n",
    "            test = pred[i] * test_label[i]\n",
    "            res.append(test)\n",
    "        \n",
    "        correct = len([i for i in res if i > 0])\n",
    "        test_accuracy = correct / m\n",
    "    \n",
    "        return test_accuracy, pred"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "81bec920",
   "metadata": {},
   "source": [
    "### Fit model and use test data to test the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "960094c5",
   "metadata": {},
   "outputs": [],
   "source": [
    "svm = binary_SVM(C = 10)\n",
    "svm.fit(train_data[0:5000], train_label[0:5000])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "9c8e4498",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test accuracy: 0.7572\n"
     ]
    }
   ],
   "source": [
    "n_test = 5000\n",
    "test_accuracy, pred = svm.predict(test_data[0:n_test], test_label[0:n_test])\n",
    "print(\"Test accuracy: {}\".format(test_accuracy))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a5d1243c",
   "metadata": {},
   "source": [
    "### Plot the confusion matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "f10d7926",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "conf_matrix = confusion_matrix(y_true = test_label[0:n_test].reshape(-1,), y_pred=np.round(pred.reshape(-1,)))\n",
    "df_cmt = pd.DataFrame(conf_matrix, index = ['Healthy', 'Covid'], columns = ['Healthy', 'Covid'])\n",
    "plt.figure(figsize = (10,7))\n",
    "sn.heatmap(df_cmt, annot=True, cmap=\"YlGnBu\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fbc2ddcb",
   "metadata": {},
   "source": [
    "### Plot roc curve with sklearn.svm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "18159d36",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAEWCAYAAAB42tAoAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAABAm0lEQVR4nO3deZxN9f/A8dfbjN3Y963Ivi+DkELZiqKIFNl+vr5CaA+llFRCKtmjbN/KVkpFFFliMBi7LGPs6zDGMsv798e5NE2zXMydO8v7+Xjcx9xzz+ec875nZs77fD6fcz5HVBVjjDEmPhm8HYAxxpiUzRKFMcaYBFmiMMYYkyBLFMYYYxJkicIYY0yCLFEYY4xJkCUKk26ISFYR+V5EQkXkm/Qeh7tE5GkR+cXbcRjvsUSRRonIIRG5IiJhInJCRGaISI5YZRqIyAoRueQ6aH0vIpVilckpIuNEJNi1rv2u6fzxbFdEZICIBInIZREJEZFvRKSqJ7+vm9oDhYB8qtohpcTh+t2848V4EqSqs1W1+e0sKyK5RWS662/wkojsFZFXXPN2i0iPOJZ5XkQCXO9/ExEVkeqxyixyfd74duIyt8YSRdrWRlVzADWAmsBrN2aISH3gF2AxUBQoBWwF1ohIaVeZTMCvQGWgJZATaACcBerGs82PgeeBAUBeoBywCHjkVoMXEd9bXSYRdwF7VTXSy7HcdhxJzZXYPXkcGAvkACoCuYBHgb9c82YCXeNYpotr3g17Y5YTkXzAvcBpD8Rr4qKq9kqDL+AQ8FCM6Q+AH2JMrwYmxLHcUuBL1/tewEkgh5vbLAtEAXUTKPMb0CvGdDfgjxjTCjwH7AMOAhOB0bHWsRgY7HpfFJiPc9A4CAyIZ7tvAdeBCCAM6IlzojQUOAycAr4EcrnK3+2KpScQDKyKY535gSXABeCca59mcM2r6PquF4AdwKPxxPEf1/vrrunv44lfgb6u/XIJGAHcA6wDLgJfA5lcZfO44joNnHe9Lx7rd/AusAa4ApQBmgN7gFBgAvD7jd9TPL+jPq5YzgOfARJP3EFA23jmFQcigbtifFbRtS/yx4j1DSAE8HF91g/43PVZY2//r6WHl9cDsJeHfrExEoXrH3I78LFrOhvOAb1JHMt1B4673s8DZt7CNvsAhxMp8xuJJ4plOLWRrMD9wJEbByLXQfAKToLIAGxyHUgyAaWBA0CLeLY9HJgVY7oHsN+1XA5gAfCVa97drli+BLIDWeNY33s4iSyj69UIENf7/cDrrria4hzcy8cTxwzgnUT2mwLf4dTqKgPXcGp7pXHO1HcCz7rK5gOecP2e/YBvgEWxfgfBrvX4AgVwks3jrunncZJXQoliCZAbKImTkFrGE/dUnETZHSgbx/xlwNBY+zR2rL1war+tXJ9tAOpjiSLZXtb0lLYtEpFLOAfaU8Cbrs/z4hxkj8exzHGcM2VwDjhxlYnPrZaPz3uqek5Vr+CcpSvOQRic9v11qnoMqAMUUNW3VfW6qh4ApgCd3NzO08AYVT2gqmE4TXOdYjUzDVfVy65YYosAiuCcEUeo6mp1jmT34iSeUa64VuAcWJ+6lZ0Qh/dV9aKq7sA5U//FFXsoTk2wJoCqnlXV+aoarqqXcGoPD8Ra1wxV3aFO81crYIeqLnBNjwdOJBLLKFW9oKrBwEqc5s249Adm49QCdrr6uFrFmD8Tp6kJVxPY0/yz2emGL4GuIlIeyK2q6xKJzyQhSxRpW1tV9QMaAxX4OwGcB6JxDnKxFQHOuN6fjadMfG61fHyO3HjjOvDO4++DbGecAw84bf1FReTCjRfOWXwhN7dTFKfZ6YbDOGfUMZc/Qvw+xKk5/CIiB0Tk1RjrPaKq0bHWXczNuOJzMsb7K3FM5wAQkWwiMklEDovIRWAVkFtEfGKUj/m9ivLvfR6SSCwxE0n4jW3HpqpXVHWkqtbGOZH4GvhGRPK6iiwAiojIvTh/p9mAH+JY1QKcmll/4KtEYjNJzBJFOqCqv+M0b4x2TV/GaduO68qfJ3GaNACWAy1EJLubm/oVKC4i/gmUuYxzMLihcFwhx5qeC7QXkbuAejh9EuAc3A6qau4YLz9VfdjNeI/hJJsbSuK0mcc8AMc7vLKqXlLVF1S1NNAGGCwiD7rWWyJWJ3FJ4Gh8q3IzXne9AJQH6qlqTpzmO3CaxeLa5nGc5kmnkIjEnE4qqnoRGInTlFfK9Vk48C1OZ3UXYJ6qXo9j2XCcWtN/sUSR7CxRpB/jgGYiUsM1/SrwrOtSVj8RyeO6RLM+TocrOP+QR4D5IlJBRDKISD4ReV1E/nUwVtV9OB2hc0WksYhkEpEsItIpxtl2IPC466y3DE5ncYJUdQtOO/hU4GdVveCatQG4KCKvuO5N8BGRKiJSx819MhcYJCKlXJcOjwT+p25ejSQirUWkjOvAehGn3ycK+BMnIb4sIhldl3C2wakZxeUkTl9DUvHDqWFccJ25v5lI+R+AqiLS1tXs9hxxJ/BbJiLDRKTOjb8FnP6PCzgd5zfMBDri9KvE1ex0w+vAA6p6KCliM+6zRJFOqOppnHbeYa7pP4AWOB2Yx3GaRmoC97kO+KjqNeAhYDdOp+NFnINzfpyDYVwGAJ/iXAlzAedSyHbA9675Y3GuajmJc1CY/e9VxGmuK5Y5Mb5TFM4BuAbOFU9ncJJJLjfXOR0nGa5yLX8Vp2nDXWVxal1hODW0Car6m+uM+FGctv8zOMmzq6rujmc904BKruazRbew/fiMw7kQ4AywHvgpocKqegandvkBTvNhJSAAp8P8TinwhSuWY0Az4BFXn9ANq3CutjqqqhsTiPOY6+/WJLMbV5IYYwxws1M5BHhaVVd6Ox7jfVajMMYgIi1cd1FnxmniEZzaiDGeSxSu2/ZPiUhQPPNFRMa7LpfbJiK1PBWLMSZR9XGaCc/gNOe1jeeSYJMOeazpSUTux2m7/VJVq8Qx/2Gc9uCHca5k+VhV63kkGGOMMbfNYzUKVV2FM6xBfB7DSSKqqutxrvNOimvwjTHGJKGkHnTtVhTjnzf9hLg++9edvSLSG+gNkD179toVKlRIlgCNMSY1i1Zl/6EQLp0/DapnVLXA7azHm4lC4vgsznYwVZ0MTAbw9/fXgIAAT8ZljDGplqqy5cgFvgk4wpJtx8myfQ3lruxj09J5hxNfOm7eTBQhQIkY08VxrrM2xhhzG7aFXOCNr/9k+YyPyJqvCF36DOKZHoOoVTIPIvHd75k4byaK74B+4kRfDwhV1aQYUM4YY9KVQ2cuM275XuZ8/S3nl00k+koo/V99jfc61kiS9XssUYjIXJxBvvKLSAjOMAIZAVR1IvAjzhVP+3EGFevuqViMMSatORF6le+3HmPJtmNs3nOIC79OImzXH1SrXp0vpi+jVq2ku+PAY4lCVRMcUtk1QuVzSbGtiIgIQkJCuHr1alKszhi3ZMmSheLFi5MxY0Zvh2LSiahoZcXuU8zbEMzKPaeIVqhaLBddquRg3PTNvPvuu7z00ktJ/jfpzaanJBMSEoKfnx933303zvhsxniWqnL27FlCQkIoVaqUt8Mx6UBoeAR9Zm1i3YGzFPDLTKeKWch8LJA3+78AwAvtg8mXL59Htp0mhvC4evUq+fLlsyRhko2IkC9fPqvFmmRx6Mxl2k1Yw6bD53m3bWW65NzNJ8+1ZfS7wzl+3Ona9VSSgDSSKABLEibZ2d+cSQ5bgs/TbsIazodf592meZn4UheeH9Cfhg0bEhQURJEinr9POU00PRljTFp0+Oxlen+1Cb8sGZnUqQr3165EVFQUM2bMoGvXrsl2smKJwhhjUpgr16N4/6fdzP7zMHrhGF8O7UDFIrn46quvqFGjBoULJ8lzpdyWZpqevM3Hx4caNWpQpUoV2rRpw4ULF27O27FjB02bNqVcuXKULVuWESNGEHMwxqVLl+Lv70/FihWpUKECL774ohe+QcK2bNlCr169vB1GvK5du0bHjh0pU6YM9erV49ChQ3GWa9y4MeXLl6dGjRrUqFGDU6dOJbj86dOnadmyZTJ9C2Mco5bu4ovVeymweyGHJvVh06/Oc79atmyZ7EkCLFEkmaxZsxIYGEhQUBB58+bls88+A+DKlSs8+uijvPrqq+zdu5etW7eydu1aJkyYAEBQUBD9+vVj1qxZ7Nq1i6CgIEqXTsqnYkJkpFtP9kzQyJEj6d/f/Ye/JcU2b8W0adPIkycP+/fvZ9CgQbzyyivxlp09ezaBgYEEBgZSsGDBBJcvUKAARYoUYc2aNcnyPUz6pqpMXX2Ayd8u5fLcF1i3YCpdu3blkUce8Wpcaa7p6a3vd7Dz2MUkXWelojl5s01lt8vXr1+fbdu2ATBnzhwaNmxI8+bNAciWLRuffvopjRs35rnnnuODDz5gyJAh3Bjo0NfXl759+/5rnWFhYfTv35+AgABEhDfffJMnnniCHDlyEBbmPFXy22+/ZcmSJcyYMYNu3bqRN29etmzZQo0aNVi4cCGBgYHkzp0bgDJlyrBmzRoyZMhAnz59CA4OBmDcuHE0bNjwH9u+dOkS27Zto3r16gBs2LCBgQMHcuXKFbJmzcoXX3xB+fLlmTFjBj/88ANXr17l8uXLfP/99/Tv35/t27cTGRnJ8OHDeeyxxzh06BBdunTh8uXLAHz66ac0aNDA7f0bl8WLFzN8+HAA2rdvT79+/VBVt9twE1q+bdu2zJ49+1/7xZikdO7ydd5YHMSsz8cQumYOJUuWZM7PP988dnhTmksU3hYVFcWvv/5Kz549AafZqXbt2v8oc8899xAWFsbFixcJCgrihRdeSHS9I0aMIFeuXGzfvh2A8+fPJ7rM3r17Wb58OT4+PkRHR7Nw4UK6d+/On3/+yd13302hQoXo3LkzgwYN4r777iM4OJgWLVqwa9euf6wnICCAKlX+fqRIhQoVWLVqFb6+vixfvpzXX3+d+fPnA7Bu3Tq2bdtG3rx5ef3112natCnTp0/nwoUL1K1bl4ceeoiCBQuybNkysmTJwr59+3jqqaeIa6DHRo0acenSpX99Pnr0aB566KF/fHb06FFKlHCGDvP19SVXrlycPXuW/Pnz/2v57t274+PjwxNPPMHQoUMRkQSX9/f3Z+jQoYnub2Nuh6ry3dZjDP9uB2HXIunUshGZauZl5MiR5MiRw9vhAWkwUdzKmX9SunLlCjVq1ODQoUPUrl2bZs2aASR4VnsrVywsX76cefP+HtQrT548iS7ToUMHfHx8AOjYsSNvv/023bt3Z968eXTs2PHmenfu3HlzmYsXL3Lp0iX8/Pxufnb8+HEKFPh7dOLQ0FCeffZZ9u3bh4gQERFxc16zZs3ImzcvAL/88gvfffcdo0ePBpz7XYKDgylatCj9+vUjMDAQHx8f9u7dG2f8q1evTvQ73hDXA7ji2r+zZ8+mWLFiXLp0iSeeeIKvvvqKrl27Jrh8wYIFOXbMxqs0Sevc5evM3xTCl78FseWbjyl+V2mWTB1D+cJ+iS+czNJcovCWG30UoaGhtG7dms8++4wBAwZQuXJlVq1a9Y+yBw4cIEeOHPj5+VG5cmU2bdp0s1knPvElnJifxb75K3v27Dff169fn/3793P69GkWLVp08ww5OjqadevWkTVr1gS/W8x1Dxs2jCZNmrBw4UIOHTpE48aN49ymqjJ//nzKly//j/UNHz6cQoUKsXXrVqKjo8mSJUuc272VGkXx4sU5cuQIxYsXJzIyktDQ0JsJK6ZixYoB4OfnR+fOndmwYQNdu3ZNcPmrV68muH+McZeqsu7AWeZuOMLPQSc4v2MVl1ZMIjL8Ep3a358ikwRYZ3aSy5UrF+PHj2f06NFERETw9NNP88cff7B8+XLAqXkMGDCAl19+GYCXXnqJkSNH3jyrjo6OZsyYMf9ab/Pmzfn0009vTt9oeipUqBC7du262bQUHxGhXbt2DB48mIoVK968izP2egMDA/+1bMWKFdm/f//N6dDQ0JsH3BkzZsS7zRYtWvDJJ5/cPFvfsmXLzeWLFClChgwZ+Oqrr4iKiopz+dWrV9/sdI75ip0kAB599FFmzpwJOH01TZs2/VdijYyM5MyZM4AzPtiSJUtuNqkltPzevXv/0fRmzK26dDWCqasP0PSj3+k85U+WB+zCd+UYziweRZVypdm0KYDhb77p7TDjZYnCA2rWrEn16tWZN28eWbNmZfHixbzzzjuUL1+eqlWrUqdOHfr16wdAtWrVGDduHE899RQVK1akSpUqN2/Jj2no0KGcP3+eKlWqUL16dVauXAnAqFGjaN26NU2bNk30Ds2OHTsya9asm81OAOPHjycgIIBq1apRqVIlJk6c+K/lKlSoQGho6M2z+5dffpnXXnuNhg0bxnuQB6fmERERQbVq1ahSpQrDhg0DoG/fvsycOZN7772XvXv3/qMWcrt69uzJ2bNnKVOmDGPGjGHUqFE359WoUQNwLoFt0aIF1apVo0aNGhQrVoz/+7//S3T5lStXev2qE5M6HTkXzntLd9HgvRW888Mu8mXPxJgnqzPtyTIEb1vH+++/z/r16xNtUfA2iattNiWL6wl3u3btomLFil6KKH0YO3Ysfn5+KfpeCk+5//77Wbx4cZz9Qva3Z2ILDY/gi7UH+WXHSXYev0gGgYerFqF1KV/2Bfx+8zLz8+fPu9XXmFREZJOq+t/OslajMG7573//S+bMmb0dRrI7ffo0gwcPTtZ/aJN6XY2IotuMDXz86z6yZ/ZhyMMVWfnC/ZQ78wftmzVgyJAhnDhxAnDvgpSUIs10Zt/KNfPm1mXJkoUuXbp4O4xkV6BAAdq2bRvnvNRWGzeeExWtfLf1KB8v38fhc+F8/nRtWlYpzK5du+j8WEvWrl1Ly5YtmTRpklfurL5TaSJRZMmShbNnz9pQ4ybZ3HgeRXxXbJn0Y0vweV6Zv429J8OoWCQnX3SrQ+PyBQkPD+f+++8nOjqaL7/8kmeeeSbVHp/SRKIoXrw4ISEhnD592tuhmHTkxhPuTPoUHa18tGwPn//2F4VzZuHzp2vRonJh9u7dg2oBsmXLxuzZs6levTqFChXydrh3JE0kiowZM9pTxowxyWr+5hA+W/kX7WsX5402lciokbz22quMHj2amTNn8swzz6SI4TeSQppIFMYYk5wuhF/nkxX7qVIsJx+2r8bq1avp1asX+/bto1evXrRu3drbISYpu+rJGGNuwdELV2g/cR0nQq/y+sMVefvtt3nggQeIjIxk+fLlTJky5ebgm2mFJQpjjHHT7hMXeWLCWk6GXmVG9zo0uMcZNHLQoEFs376dBx980NsheoQlCmOMSYSq8nXAEdp9tpbrly9QascXLJ/jPFPmkUceYcyYMUkywkBKZX0UxhgTD1Vl1b4zfLZyP38eOEux84HsXTie7RfOU7dG+hn/yxKFMcbEculqBEuDTjBz7SF2HLtIXsLIt24a61Yvw9/fn2nTllOtWjVvh5lsLFEYYwxO7WHtX2f5OuAIP+84wdWIaO4pkJ0PnqjGXXqSZhPW8+GHHzJw4EB8fdPXoTN9fVtjjIlD8Nlwhi0O4ve9p8mVNSPtaxenbv5I/tr4G0/WaQyUIDg4OM1dzeQu68w2xqR7fedsYvPh8wxrXYl1rzYm38HldGrekOHDh98cxC+9JgmwRGGMSed2HAsl6OhFXmhejnvzhNP0gfsZPHgwTZs2ZceOHalyEL+kZk1Pxph07dMV+8mR2Zfm5fNQrUIdRIQ5c+bQqVOnVDuIX1KzRGGMSbfW/nWGxb9t4IUnm1I0f27mzZtH9erVKVCggLdDS1Gs6ckYky6tCArm8W59OT69H7mPrgfgoYcesiQRB0sUxph05VpkFL1GzaTF/fdy8o9v6PhMN9o/3tbbYaVo1vRkjEkXVJV1B87S9bmX+OvnGeQpXILZPy+jVfOHvB1aimeJwhiTpp26eJVvN4fwvw3BHD53hYwF7qFDtz7M+OwjsmXL5u3wUgWPJgoRaQl8DPgAU1V1VKz5uYBZQElXLKNV9QtPxmSMSR8uXo1g2KIgFq/fzelfJlHqnjKMfXs4raq0JEtGH2+Hl6p4LFGIiA/wGdAMCAE2ish3qrozRrHngJ2q2kZECgB7RGS2ql73VFzGmLTveOgVnpn6J0GrfiTstylEXLnM011a0K6mPbr2dniyRlEX2K+qBwBEZB7wGBAzUSjgJ87FyjmAc0CkB2MyxqRxYdci6TzuRwJmf0jYvj+pV68e06ZNo3Llyt4OLdXy5FVPxYAjMaZDXJ/F9ClQETgGbAeeV9Xo2CsSkd4iEiAiAadPn/ZUvMaYVC4yKpr+czaz7/Ax9PhOxowZw5o1ayxJ3CFPJoq4bmnUWNMtgECgKFAD+FREcv5rIdXJquqvqv52jbMxJi579+7j4f97lZV7TvP+fx7l2NEQBg0ahI+P9UfcKU8mihCgRIzp4jg1h5i6AwvUsR84CFTwYEzGmDQmIiKCF994h4pVqrB8zgQ6V83J0/XuImfOf51zmtvkyUSxESgrIqVEJBPQCfguVplg4EEAESkElAcOeDAmY0wasfXIBfp/toh891TloxHDyF66Fp8vWMm7ne/zdmhpjsc6s1U1UkT6AT/jXB47XVV3iEgf1/yJwAhghohsx2mqekVVz3gqJmNM6hcVrYxZtodPftnB0c97kMnXh34jPuHN53uR3y+Lt8NLk0Q1drdByubv768BAQHeDsMY4wWXr0XS6f2v2XY5J0/VLUmDbMdpUKc2+fPn93ZoKZ6IbFJV/9tZ1u7MNsakCpcvX6ZVl+dYvfBL+r45hlFPtAbSz3OrvckGBTTGpGjXI6MZNe0bipQqz+qFM6narD3vDe7h7bDSFUsUxpgUR1UJOHSOIQu3U+LBLrzW60muR0O3kTP4feEsu6IpmVnTkzEmxbh0NYIpqw+ycEsIwWcvkzWTL/XurU/2cnmZNHYUOXNk93aI6ZIlCmNMirA5+DwD5wVyKOQYGdbPoF6NKkz/5ENyZG7l7dDSPUsUxhivm/3nYYYtCsL34B9c/HkSV8IvU/uJh8iR2Q5RKYH9FowxXhMVrYxauovPf9iArJnCwW1rqV+/PlOnTqVSpUreDs+4WGe2McZr3lgcxJTVB3mkbA4uHd7Bxx9/zOrVqy1JpDBWozDGeMXUJX8wYfwXvPDCCwxtXYnRvVrg5+fn7bBMHKxGYYxJVuFXr/NEn1fo3bYpl//8mi41cgNYkkjBLFEYY5LF1Ygo3prxIwXvqcKCSR9QonpDdu3cyV3Fi3o7NJMISxTGGI/bc+ISLUcv4+2+TxEZdpa3xk/jUMAK7rm7ROILG6+zPgpjjEeN/d8yJm6PwC9LJj6aOIOurR8gb9683g7L3AJLFMYYjwgLC6Nt9/78+u1M6nYbyvfjX6egDQOeKlmiMMYkuWXLlvFU1x6cPRFClWZP8uOYF8hnSSLVsj4KY0ySev3112nevDmh16HD8OlsWTqXfHlyeTsscwfcrlGISHZVvezJYIwxqVd0dDTnwiPYeKUgOe/twH8GvsyoDrXJkEG8HZq5Q4kmChFpAEwFcgAlRaQ68B9V7evp4IwxKd+JEyf4b9/niPArypHSbbiSvTyfjGlLl3vvQsSSRFrgTtPTWKAFcBZAVbcC93syKGNMyqeqTJwyjTLlKrD4++9ZExxOjRK5WdL/PrrWv9uSRBriVtOTqh6J9UuP8kw4xpjUYPue/bTr9Cx/Ba4lc/FKNBk4lBFdm9GgjD27Oi1yJ1EccTU/qYhkAgYAuzwbljEmpVq+8ySDPv+Vg7u30aTHq4x+4yVqlsxjNYg0zJ1E0Qf4GCgGhAC/ANY/YUw6s37zdl78aBohJZpR4Z6KTA3cTb3yxbwdlkkG7iSK8qr6dMwPRKQhsMYzIRljUpIDJ0P5vxffZOXcCUimLLw1sxMvt6tLJl+7uj69cOc3/Ymbnxlj0pjZP/xOpeq1WTHrY8rWeYC1GwMZ2uFeSxLpTLw1ChGpDzQACojI4BizcgI+ng7MGONdWw6coNuTjyI+vkz5cg69ujzl7ZCMlyTU9JQJ594JXyDmQPEXgfaeDMoY4z2bN29mT0Q+hn+/k9Kd3uDLl9pTr8Jd3g7LeFG8iUJVfwd+F5EZqno4GWMyxnjBpUuXeG7gi3w1fTL5HhlEy7YdGftKPwr4ZfZ2aMbL3OnMDheRD4HKwM1RvVS1qceiMsYkq6VLl9Kley/OnjxOvnpteW9QT3o0qYSPDb9hcK8zezawGygFvAUcAjZ6MCZjTDIa/OLLPPzww1yM9KHV61PY8ctc/u/BypYkzE3uJIp8qjoNiFDV31W1B3Cvh+MyxnhYVFQUR86Fsya8EHkbdmLi/GX88E4PCuW04cDNP7nT9BTh+nlcRB4BjgHFPReSMcaTjh8/TteevbmYpTBnKrTDN28l5k/rQuPyBb0dmkmh3EkU74hILuAFnPsncgIDPRmUMSbpqSqfTJzCyy++yPXr1yjY5FkG3FeKHg1LUTiX1SJM/BJNFKq6xPU2FGgCN+/MNsakEocOHeLxp55ly/pVZClRmVfe+ohXOzYlV7aM3g7NpAIJ3XDnAzyJM8bTT6oaJCKtgdeBrEDN5AnRGHMnoqOVr37fydbALVR9cjDffvwG5QrbE+eM+xKqUUwDSgAbgPEichioD7yqqovcWbmItMQZUNAHmKqqo+Io0xgYB2QEzqjqA+6Hb4yJz44dO/j4i3kcu6s5QUeVp8Z+z5Qe95E1kw2sYG5NQonCH6imqtEikgU4A5RR1RPurNhVI/kMaIYz6uxGEflOVXfGKJMbmAC0VNVgEbHeNGPu0PXr13n+9beY/PGHkDErtV+owUcd6tG2ZjG75NXcloQSxXVVjQZQ1asistfdJOFSF9ivqgcARGQe8BiwM0aZzsACVQ12befULUVvjPmHgIAA2jz5DCcO7iFP1ca89d6H9G5Rk8y+Voswty+hRFFBRLa53gtwj2taAFXVaomsuxhwJMZ0CFAvVplyQEYR+Q1nPKmPVfXL2CsSkd5Ab4CSJUsmsllj0qfT50O5v+lDXFcfHnt5HHNGPEe2TG49xNKYBCX0V1TxDtcdVx1X49h+beBBnA7ydSKyXlX3/mMh1cnAZAB/f//Y6zAmXVu4bDV/RRdg4ZZj5GzzGk82b8AHnRuQJaPVIkzSSGhQwDsdCDAEpzP8huI4N+vFLnNGVS8Dl0VkFVAd2IsxJl77T4Uxf90+Pv3gLULWLib/I4N46NEnGTOiF/VK5/N2eCaN8WS9dCNQVkRKAUeBTjh9EjEtBj4VEV+cYc3rAWM9GJMxqVJkVDRbjlxgxe5TrNh1isC1Kzj782dEhZ2jZaeeTBg7hFKFLUEYz/BYolDVSBHpB/yMc3nsdFXdISJ9XPMnquouEfkJ2AZE41xCG+SpmIxJjX7ddZKXvt3GucvX8c0gZNw0l1O/zKJchYp8OWMJ9erF7vozJmmJauJN/iKSFSipqns8H1LC/P39NSAgwNthGONxkVHRfLHmEO8t3UXFIn78p1FpGlcsxPpVK1mzZg2vv/46mTPbsyKMe0Rkk6r6386yidYoRKQNMBqnaaiUiNQA3lbVR29ng8aYhKkq3287zthlezl45jINCgsXl41mw6lqPPrOOzRv3pzmzZt7O0yTjrgzzPhwnHsiLgCoaiBwt6cCMiY9O3IunC7TNjBg7hYy+Qjtsu7mhzef4tfly8ifP7+3wzPplDt9FJGqGipid3Qa4ymHz17m64AjTP/jEBkEBtTNxZJP32DcypU0btyYKVOmUKZMGW+HadIpdxJFkIh0BnxEpCwwAFjr2bCMST++23qMgfO2ANC8UmHeaFOJs0f2M3LbNiZNmkSvXr3IkMGdyr8xnuFOougPDAGuAXNwrmJ6x5NBGZMeREUr8zeHMHRREP535aVv9UysXfkzRXPXpmjuqgQHB5MtWzZvh2mMW4mivKoOwUkWxpgksOnwOYYsDGL3iUtUL5qd0keW0nzAKHLlykWvXr0oWLCgJQmTYrhTnx0jIrtFZISIVPZ4RMakYdHRylfrDtFp8nrCrkXyfLUMHJjSn/dHjqBDhw7s3LmTggVtEGWTsrjzhLsmIlIY5yFGk0UkJ/A/VbXmJ2Nuwf5TYQxbFMS6A2e5v1wBRrUpR9UK95A1a1a+++472rRp4+0QjYmTW3dmu4YXHy8iK4GXgTewfgpj3BJ0NJRPV+zn550nyJHJl/+rqLz6TG18fHxYvHgxVatWJVcue+KcSbkSbXoSkYoiMlxEgoBPca54Ku7xyIxJ5UKvRPDq/G20+fQP1v51hh51ClHryDcM7daG2bNnA3DfffdZkjApnjs1ii+AuUBzVY09+qsxJg7XIqPoNXMjW4Iv0LNhKcpe28Pg/j04ceIEL774Iu3bt/d2iMa4zZ0+inuTIxBj0oIL4ddZsPkoczcEs+9UGOOfqsnqWWPpNHo0VatWZdGiRdSpU8fbYRpzS+JNFCLytao+KSLb+ecDh9x9wp0x6YKqsu7AWeZtOMJPO05wPTKaasVyMq5DVR6tXpSsp5qTM2dOXnnlFTJlyuTtcI25ZQnVKJ53/WydHIEYkxpFRSuDvw5kceAxcmbx5ak6JXigmA9j33qZjSeq0bb2uzRr1oxmzZp5O1Rjblu8ndmqetz1tq+qHo75AvomT3jGpFwRUdG8On8biwOP8fyDZVn/WlOKnlhD26b1WLFiBYULF/Z2iMYkCXduuIvrVKhVUgdiTGqyOfg8bT75g282hfD8g2V57B5fHm7RjD59+lC3bl22b99O//79vR2mMUkioT6K/+LUHEqLyLYYs/yANZ4OzJiUKDpaeffHXUxfc5BCflmY+EwtWlYpwvbt29m5cydTp06lR48e2GjLJi1JqI9iDrAUeA94Ncbnl1T1nEejMiaF+vCXPUz74yBP1ytJmxIRLF80jZZVhlK1alUOHz5M1qxZvR2iMUkuoaYnVdVDwHPApRgvRCSv50MzJmVZsfskn//2Fx1qFibD5q9pVL8e48eP59SpUwCWJEyalViNojWwCefy2Jh1aQVKezAuY1KUc5ev8/K32yl8NZgf3n6ZXTt30qVLF8aOHUu+fPm8HZ4xHhVvolDV1q6fpZIvHGNSnoioaPrP3cz50IucnTmEnH45+PHHH2nVyq7pMOlDondmi0hDIFBVL4vIM0AtYJyqBns8OmO87NSlq3R/fw5BEQUY3bEOxdotoWrVqvj5+Xk7NGOSjTuXx34OhItIdZyRYw8DX3k0KmO8TFX5Zs1uKjZux4/v9qRl5r108C9BgwYNLEmYdMedRBGpqgo8Bnysqh/jXCJrTJq089hFHuj3AU+1bMj5wGX07jeIca/ZPaYm/XJn9NhLIvIa0AVoJCI+QEbPhmVM8gu/HsnIH3fxychhXNy4mBJlKvLN7J+oV9cG8TPpmzuJoiPQGeihqidEpCTwoWfDMiZ5bQ+5QP85mzh8/iqPtW7NXQ/X4o0hr5Exo50TGePOMOMnRGQ2UEdEWgMbVPVLz4dmTPKY/vNGBjzXl1zFy/HN1I9pUCa/t0MyJkVx5wl3TwIbgA44z83+U0TsqSsm1Qu7ep0WvYfQ69EHuHokiAGP1rUkYUwc3Gl6GgLUUdVTACJSAFgOfOvJwIzxlN0nLjJh8Rqmv/cylw8HUaZmA5Z8/SXly9zj7dCMSZHcSRQZbiQJl7O4d7WUMSlGdLTy/bZjTP/jIFtDQtGzwURfOM6bH37Kmy/0tUH8jEmAO4niJxH5Gee52eB0bv/ouZCMSVqr9p5m5I+72LZ1K1mObWbYsDdoV7MZ2T7pTpYsWbwdnjEpnjud2S+JyOPAfTjjPU1W1YUej8yYJLDz2EW6TPkD3TSfk6vmUSB/ftqUe5+82e2RpMa4K6HnUZQFRgP3ANuBF1X1aHIFZsydUlVemfANJ2e8xbUzR3j22WcZM2YMefPa4MfG3IqE+hqmA0uAJ3BGkP0kWSIyJol8sGQrP340iOw+Ufz000/MmDHDkoQxtyGhpic/VZ3ier9HRDYnR0DG3Kl169ZRtGxVpq4/xhOvjWfqwMfJmTOnt8MyJtVKqEaRRURqikgtEakFZI01nSgRaSkie0Rkv4i8mkC5OiISZfdnmDtx/vx5unXrToMGDWj8n7fJIMLHz3e0JGHMHUqoRnEcGBNj+kSMaQWaJrRi15hQnwHNgBBgo4h8p6o74yj3PvDzrYVuzN+++fZbevfpy4XzZ8l5bwdqNX6YoY9Vp0gue+qcMXcqoQcXNbnDddcF9qvqAQARmYczAu3OWOX6A/MBG3nN3JanevRh3heTyFiwNI1eGMWIXm24v2x+uzfCmCTizn0Ut6sYcCTGdAhQL2YBESkGtMOpncSbKESkN9AboGTJkkkeqEl9VJVDpy/xyW8HWBFegrta9OSTkW/QumYJSxDGJDFPJoq4/ls11vQ44BVVjUron1tVJwOTAfz9/WOvw6QzawN38XS3nlzIVpwCTbsz6NkneP7BsmTN5OPt0IxJkzyZKEKAEjGmiwPHYpXxB+a5kkR+4GERiVTVRR6My6RSR85dpsfL7/Drl+NAhFY9HmLaS00onMvurjbGk9x5ZrYATwOlVfVt1/MoCqvqhkQW3QiUFZFSwFGgE85zLW5S1VIxtjMDWGJJwsR28WoEb321nAlvv8jVkJ3cU7Mhc2dOo07V8t4OzZh0wZ0axQQgGqcf4W3gEm50PqtqpIj0w7mayQeYrqo7RKSPa/7EOwncpH2qytwNR/jolz2cOHwQn7BTfPz5VPr/p4f1QxiTjMR5HHYCBUQ2q2otEdmiqjVdn21V1erJEmEs/v7+GhAQ4I1Nm2QUERVN77HfsnDRIpo/059hrStRrkAWMmfO7O3QjEmVRGSTqvrfzrLu1CgiXPc6qGtjBXBqGMYkOVVlWVAI/xn4KodWzsMvd14+fXwcBQvm8nZoxqRb7jxXYjywECgoIu8CfwAjPRqVSXfCrkUyc+0h6vT/lEca1+fQijk0f+xJDv+1l4IFC3o7PGPSNXeGGZ8tIpuAB3EueW2rqrs8HplJF1SVBZuP8t7S3Zw6d4ETXwwlT+5cfPHjUh5p1dLb4RljcO+qp5JAOPB9zM9UNdiTgZm0L/hsOC9+u5VVq1dTr159JnVpSsSTP1OlShVy5Mjh7fCMMS7u9FH8gNM/IUAWoBSwB6jswbhMGhYaHsHsDYf5eMlmTi2bzIVtv9L2wS+ofVcjuOteb4dnjInFnaanqjGnXSPH/sdjEZk0K/hsONPXHOR/G4M5s30VYSsnERl+iWHDhtG581PeDs8YE49bvjNbVTeLiA3gZ9ymqny57jAjluxEBPy2zObM0jnUrl2badOmUb26V660Nsa4yZ0+isExJjMAtYDTHovIpClXI6IYsjCIbzcdoWnZfLzXoSY7N2UhoHF1Bg8ejK+vJ0eRMcYkBXf+S/1ivI/E6bOY75lwTFoSdDSUVxdsY8uOveTYOJ2C3EfhXPUp3LQpTZsm+DgTY0wKkmCicN1ol0NVX0qmeEwacOlqBO8s2cX/Nh4icvtPnPttBpd8fbnnP896OzRjzG2IN1GIiK9rvCa3HntqzA3DFgUxf2UAsmoCR3cH0qpVKyZNmkSJEiUSX9gYk+IkVKPYgNMfESgi3wHfAJdvzFTVBR6OzaQy0dHKzHWHWBR4jM51ijFrySlmzZpF586dbRA/Y1Ixd/oo8gJncUaPvXE/hQKWKMxNR86F03P0PDb89gtteg5ixLN1eLvLARvEz5g0IKFEUdB1xVMQfyeIG+wpc+amK1eu8FDH/2Pvr/PIk78AH7UZTybfDOBrScKYtCChQQF9gByul1+M9zdexvD7779TsXIV9i6fQ6PWHTiwd7cN4mdMGpNQjeK4qr6dbJGYVCcsLIx2jz9OlG9WinUeyaLJL5A7eyZvh2WMSWIJJQrrfTRxWr16NTX86zH+92DytHuTiJzF6NWkInktSRiTJiWUKB5MtihMqnDmzBkGDhzI7NmzqdVlCOeL1eeRBxvR875S1CiR29vhGWM8JN5EoarnkjMQk3KpKl9//TX9+/fn3LnzFG3ShdAidZjwdG1aVins7fCMMR5mA+2YRD3//PN88sknFC9XlYJt3qRe7RoMbV2JWiXzeDs0Y0wysERh4qSqREREkClTJtq1a8e1LHn5iVp0rns377argq+PO0/RNcakBfbfbv7lr7/+4sEHH2To0KEAZClZjT9zNKBi0dy83bayJQlj0hn7jzc3RUVFMWbMGKpWrcqmTZsoX748i7Yc5Zlpf5I3eyYmPlObzL4+3g7TGJPMrOnJALB7926effZZNmzYQKtHHqHzoBHsvOjL6AXbqV0yD9O6+eOXJaO3wzTGeIElCgNAdHQ0IUeP0vr5UezMWoWhy46RI7MvD5QrwJuPVrIkYUw6ZokiHduwYQOLFy/mjbfe5vdTmcn17Ofsxocu9e6iVZXC1LorDxmtP8KYdM8SRToUHh7OG2+8wdixY8mTvxC/+vhz4nomHqpYhCGPVKRU/uzeDtEYk4JYokhnVq5cydPPduf4kcPkqf0w2Ro9S848efmwdSUalS3g7fCMMSmQJYp0JCwsjMfaPcEVyUKFnqNp+3AzmpQvSNMKBe2SV2NMvCxRpAO//fYb5arXZfLqw/i1HcYjdWsxo/d9dqmrMcYtdhqZhp0+fZqOnTrRpEkTaj37BrPWH+aZ1g8ytWdDSxLGGLdZjSINUlXmzp1Lv/4DCL14kVyNnqH9k0/yysNVKJE3m7fDM8akMpYo0qD+/fvz2Wefkb1ERUp3Gsn4Pm1oVbWIt8MyxqRSlijSiOjoaCIjI8mUKRN1mrSi8L5IyjZtz/Tu91KmoD251hhz+zzaRyEiLUVkj4jsF5FX45j/tIhsc73Wikh1T8aTVu3bt4+mTZsyZMgQfgo6wehtvpS4vz3/69PQkoQx5o55LFGIiA/wGdAKqAQ8JSKVYhU7CDygqtWAEcBkT8WTFkVGRjJ69GiqVatGwOYtBIRmpc+sTZQukJ0FfRtQJFdWb4dojEkDPNn0VBfYr6oHAERkHvAYsPNGAVVdG6P8eqC4B+NJU3bt2kXXrl0JCAggd8UGZG/yHy4UKsx/65Zk4ENl7aomY0yS8WSiKAYciTEdAtRLoHxPYGlcM0SkN9AboGTJkkkVX6qlquw8FsrugyHkf/QV6j/0CH2blKVphYJk8rUrno0xScuTiULi+EzjLCjSBCdR3BfXfFWdjKtZyt/fP851pAcrV/3BuGlzifLvRNDRi5TsM5Vuje7h+QfLWYIwxniMJxNFCFAixnRx4FjsQiJSDZgKtFLVsx6MJ9UK2H+c5wa/zIbvZ+Pjl59GZVowom112tUsRo7MduGaMcazPHmU2QiUFZFSwFGgE9A5ZgERKQksALqo6l4PxpLqXLkexZJtxxj35Xz+/HIUUaEnqd6iI+M/ep9GlUoiEleFzRhjkp7HEoWqRopIP+BnwAeYrqo7RKSPa/5E4A0gHzDBdeCLVFV/T8WUGuw7eYnZfwazYHMIFy5e4sS0N8mfNy/Tv/mVh5s19XZ4xph0SFRTV5O/v7+/BgQEeDuMJHfxagQjf9jFvI1HiDyyjbYPN6NL/VL4nDtI5cqVyZrVLnU1xtw+Edl0uyfi1sCdAuw/dYlnpm7g+Inj5Nwyi6A/fqZO86LUK+0PpfN5OzxjTDpnicLLTl+6xrPTN3By8y+E/jqFE+GXeffdd+ncuXPiCxtjTDKwROFFV65H0WvmRnZ+O5YLm36gfv36TJs2jYoVK3o7NGOMuckShZdER0czfFEg246GMqRfDzJfakHfvn3x8bE7qo0xKYslCi/Ys2cPPXv2Yk90IboMHMqLT9bwdkjGGBMvu503GUVERDBk+AiqVKvGn5sD0TwlaVG5sLfDMsaYBFmNIpns2LGDxzp04q9dQWQr14D7ur3CM02q07xSIW+HZowxCbJEkQxUlQVbjnHo6Cnq9HqH/703kFL5s3s7LGOMcYslCg9au3Yt8xcsJEejZ5kZdJ2e47/j487+ZMloHdbGmNTDEoUHhIWF0X/wy8ycOpGMuQpS8HpN/tuyJq+2rECGDDZGkzEmdbFEkYRUlTEzvmH4SwMJO3uCnLVb83jvF+nVtBINy+T3dnjGGHNbLFEkAVVlxe5TjPlxKz8N+z8yZs9J349m83qPthTLbWM0GWNSN0sUdyD8eiTzNx9lzBffcC5XOUrky8E7k+bSt90D5MphndXGmLTBEsVtWhx4lFe/+p3gJZ8Rvnct/d8ay0cvDiCjj92aYoxJWyxR3IbFgUf5v6GjubBiGhJ1nffee48XX+yHryUJY0waZIniFoSGRzDih51MGvkqYYE/Ub9BQ76YPo3y5ct7OzRjjPEYSxRuiI5W/rfxMO//uIOL16HL009ToUdr+j33XzJksFqEMSZts0SRiG0hFxg06QfWfPEuJStU5/svJ1K5aC5vh2WMMcnGEkUcrkZEsWznSb7dcIjvZk0idM1csufIzrBnX7MkYYxJdyxRxKCqLAo8ysgfd3P0wF5Cl44h/PhftH38CSZO+IxChWwAP2NM+mMN7C7nLl+n0+T1DPrfVormzsqHT9akUJZoFixYwML531qSMMakW1ajAELOh9NrZgA7Nv9J9ah9LBw5iQwZhI579+Lra7vIGJO+pduj4I1+iK8DjrBqRzChq2YSuukHMpYqxblzZ8mfP78lCWOMIZ0misNnL9Pti40cPHOZbCe3cfGHT7h45gQDBw7knXfeIXt2G37DGGNuSHeJYuuRC/SYsZEoVT5pX4EerbpRtGBBln63gHvvvdfb4RljTIqTLhKFqrLur7N8tf4wP+84QbZT21nw7n8pVzgXy5cvp0KFCmTOnNnbYRpjTIqU5hPFhfDrvPD1Vn7dfYrsURfJumYaO9b9yvoHilKua1eqV6/u7RCNMSZFS9OJIuR8OJ0mr+dE6BUaE8SiKe9x7do1PvjgAzp37uzt8IwxJlVIs4li1/GLDJi7hdArEVQ//A0z58zk/vvvZ+rUqZQtW9bb4RljTKqR5hLFgdNhjFiykxW7TpA9ozC1R32ijmWjWaN76d27tw3iZ4wxtyhNJYotwefpMWMj4ScPISsm0OShxjS4pw3c04hGjRp5OzxjjEmV0kSiuBYZxZRVBxi/bBfXNy3i2G+zyJkzJ/c1qOft0IwxJtVL9Ylie0goA+ZtYc+uHVxbNo5zR/bTqVMnxo8fT4ECBbwdnjHGpHqpusH+p6ATdJi0luuR0XzQoRZ5MimLFy9m7ty5liSMMSaJpMoahaoyadUB3pr8DdlPBLJo0UwK+GXmqT178PHx8XZ4xhiTpni0RiEiLUVkj4jsF5FX45gvIjLeNX+biNRyZ70vzF7PSwP7c2LOa0Qf2ohcuwRgScIYYzzAY4lCRHyAz4BWQCXgKRGpFKtYK6Cs69Ub+Dyx9R44eorxfR/l8rafGTRoEEHbt5M/f/4kjt4YY8wNnmx6qgvsV9UDACIyD3gM2BmjzGPAl6qqwHoRyS0iRVT1eHwrPX/yKIVKlmbRL9/bIH7GGJMMPJkoigFHYkyHALGvV42rTDHgH4lCRHrj1DgArp08vD+ofv36SRtt6pQfOOPtIFII2xd/s33xN9sXfyt/uwt6MlFIHJ/pbZRBVScDkwFEJEBV/e88vNTP9sXfbF/8zfbF32xf/E1EAm53WU92ZocAJWJMFweO3UYZY4wxXuTJRLERKCsipUQkE9AJ+C5Wme+Arq6rn+4FQhPqnzDGGJP8PNb0pKqRItIP+BnwAaar6g4R6eOaPxH4EXgY2A+EA93dWPVkD4WcGtm++Jvti7/Zvvib7Yu/3fa+EOeCI2OMMSZuqXoID2OMMZ5nicIYY0yCUmyi8NTwH6mRG/viadc+2CYia0UkzT4IPLF9EaNcHRGJEpH2yRlfcnJnX4hIYxEJFJEdIvJ7cseYXNz4H8klIt+LyFbXvnCnPzTVEZHpInJKRILimX97x01VTXEvnM7vv4DSQCZgK1ApVpmHgaU492LcC/zp7bi9uC8aAHlc71ul530Ro9wKnIsl2ns7bi/+XeTGGQmhpGu6oLfj9uK+eB143/W+AHAOyOTt2D2wL+4HagFB8cy/reNmSq1R3Bz+Q1WvAzeG/4jp5vAfqroeyC0iRZI70GSQ6L5Q1bWqet41uR7nfpS0yJ2/C4D+wHzgVHIGl8zc2RedgQWqGgygqml1f7izLxTwExEBcuAkisjkDdPzVHUVzneLz20dN1NqoohvaI9bLZMW3Or37IlzxpAWJbovRKQY0A6YmIxxeYM7fxflgDwi8puIbBKRrskWXfJyZ198ClTEuaF3O/C8qkYnT3gpym0dN1Pq8yiSbPiPNMDt7ykiTXASxX0ejch73NkX44BXVDXKOXlMs9zZF75AbeBBICuwTkTWq+peTweXzNzZFy2AQKApcA+wTERWq+pFD8eW0tzWcTOlJgob/uNvbn1PEakGTAVaqerZZIotubmzL/yBea4kkR94WEQiVXVRskSYfNz9HzmjqpeByyKyCqgOpLVE4c6+6A6MUqehfr+IHAQqABuSJ8QU47aOmym16cmG//hbovtCREoCC4AuafBsMaZE94WqllLVu1X1buBboG8aTBLg3v/IYqCRiPiKSDac0Zt3JXOcycGdfRGMU7NCRArhjKR6IFmjTBlu67iZImsU6rnhP1IdN/fFG0A+YILrTDpS0+CImW7ui3TBnX2hqrtE5CdgGxANTFXVOC+bTM3c/LsYAcwQke04zS+vqGqaG35cROYCjYH8IhICvAlkhDs7btoQHsYYYxKUUpuejDHGpBCWKIwxxiTIEoUxxpgEWaIwxhiTIEsUxhhjEmSJwqRIrpFfA2O87k6gbFgSbG+GiBx0bWuziNS/jXVMFZFKrvevx5q39k5jdK3nxn4Jco2GmjuR8jVE5OGk2LZJv+zyWJMiiUiYquZI6rIJrGMGsERVvxWR5sBoVa12B+u745gSW6+IzAT2quq7CZTvBvirar+kjsWkH1ajMKmCiOQQkV9dZ/vbReRfo8aKSBERWRXjjLuR6/PmIrLOtew3IpLYAXwVUMa17GDXuoJEZKDrs+wi8oPr2QZBItLR9flvIuIvIqOArK44Zrvmhbl+/i/mGb6rJvOEiPiIyIcislGc5wT8x43dsg7XgG4iUlecZ5Fscf0s77pL+W2goyuWjq7Yp7u2syWu/WjMv3h7/HR72SuuFxCFM4hbILAQZxSBnK55+XHuLL1RIw5z/XwBGOJ67wP4ucquArK7Pn8FeCOO7c3A9ewKoAPwJ86AetuB7DhDU+8AagJPAFNiLJvL9fM3nLP3mzHFKHMjxnbATNf7TDgjeWYFegNDXZ9nBgKAUnHEGRbj+30DtHRN5wR8Xe8fAua73ncDPo2x/EjgGdf73DjjPmX39u/bXin7lSKH8DAGuKKqNW5MiEhGYKSI3I8zHEUxoBBwIsYyG4HprrKLVDVQRB4AKgFrXMObZMI5E4/LhyIyFDiNMwrvg8BCdQbVQ0QWAI2An4DRIvI+TnPV6lv4XkuB8SKSGWgJrFLVK67mrmry9xP5cgFlgYOxls8qIoHA3cAmYFmM8jNFpCzOaKAZ49l+c+BREXnRNZ0FKEnaHAPKJBFLFCa1eBrnyWS1VTVCRA7hHORuUtVVrkTyCPCViHwInAeWqepTbmzjJVX99saEiDwUVyFV3SsitXHGzHlPRH5R1bfd+RKqelVEfsMZ9rojMPfG5oD+qvpzIqu4oqo1RCQXsAR4DhiPM5bRSlVt5+r4/y2e5QV4QlX3uBOvMWB9FCb1yAWcciWJJsBdsQuIyF2uMlOAaTiPhFwPNBSRG30O2USknJvbXAW0dS2THafZaLWIFAXCVXUWMNq1ndgiXDWbuMzDGYytEc5Adrh+/vfGMiJSzrXNOKlqKDAAeNG1TC7gqGt2txhFL+E0wd3wM9BfXNUrEakZ3zaMucEShUktZgP+IhKAU7vYHUeZxkCgiGzB6Uf4WFVP4xw454rINpzEUcGdDarqZpy+iw04fRZTVXULUBXY4GoCGgK8E8fik4FtNzqzY/kF59nGy9V5dCc4zxLZCWwWkSBgEonU+F2xbMUZVvsDnNrNGpz+ixtWApVudGbj1DwyumILck0bkyC7PNYYY0yCrEZhjDEmQZYojDHGJMgShTHGmARZojDGGJMgSxTGGGMSZInCGGNMgixRGGOMSdD/A7aL4ME9JK/KAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "classifier = SVM.SVC(kernel='rbf', probability=True, C=10)\n",
    "classifier.fit(train_data[0:5000], train_label[0:5000].ravel())\n",
    "probas_ = classifier.predict_proba(test_data[0:n_test])\n",
    "\n",
    "fpr, tpr, thresholds = roc_curve(test_label[0:n_test], probas_[:, 1])\n",
    "roc_auc = auc(fpr, tpr)\n",
    "\n",
    "plt.clf()\n",
    "plt.plot(fpr, tpr, label='ROC curve (area = %0.2f)' % roc_auc)\n",
    "plt.plot([0, 1], [0, 1], 'k--')\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.ylim([0.0, 1.0])\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.title('ROC Curve for soft margin SVM')\n",
    "plt.legend(loc=\"upper left\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f43c2d42",
   "metadata": {},
   "source": [
    "### Confusion matrix and ROC Curve for logistic regression"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f525d5f3",
   "metadata": {},
   "source": [
    "### Plot confusion matrix with our model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "492a19b3",
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in range(train_label.shape[0]):\n",
    "    if train_label[i][0] != -1:\n",
    "        train_label[i][0] = 1\n",
    "    else:\n",
    "        train_label[i][0] = 0\n",
    "\n",
    "for i in range(test_label.shape[0]):\n",
    "    if test_label[i][0] != -1:\n",
    "        test_label[i][0] = 1\n",
    "    else:\n",
    "        test_label[i][0] = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "e72f154f",
   "metadata": {},
   "outputs": [],
   "source": [
    "def sigmoid(X):\n",
    "    \n",
    "    sig = 1 / (1 + np.exp(-X))\n",
    "    \n",
    "    return sig\n",
    "\n",
    "def cost_function(X, Y, w, lam):\n",
    "    \n",
    "    m = X.shape[0]\n",
    "    regularized = lam * np.dot(w[1:].T, w[1:]) \n",
    "    cross_entropy = np.sum(-Y * np.log(sigmoid(np.dot(X, w))) - (1 - Y) * np.log(1 - sigmoid(np.dot(X, w))))\n",
    "    \n",
    "    res = cross_entropy / m + regularized / (2 * m)\n",
    "    \n",
    "    return res\n",
    "\n",
    "def calculate_gradients(w,X,Y,lam):\n",
    "    \n",
    "    regularized = w * (lam / X.shape[0])\n",
    "    regularized[0] = 0\n",
    "    grad = np.dot(X.T, sigmoid(np.dot(X, w)) - Y) / X.shape[0] + regularized\n",
    "    \n",
    "    return grad\n",
    "\n",
    "def update_weights(prev_weights, current_grads, learning_rate):\n",
    "    \n",
    "    weights = prev_weights - learning_rate * current_grads\n",
    "    \n",
    "    return weights\n",
    "\n",
    "def logistic_regpression_regularized(X, Y, weights, lam, learning_rate = 0.0005, num_steps = 1000):\n",
    "    \n",
    "    costs = []\n",
    "    for i in range(num_steps):\n",
    "        grad = calculate_gradients(weights, X, Y, lam)\n",
    "        weights = update_weights(weights, grad, learning_rate)\n",
    "        cost = cost_function(X, Y, weights, lam)\n",
    "        costs.append(cost[0][0])\n",
    "        \n",
    "    return weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "f51b481c",
   "metadata": {},
   "outputs": [],
   "source": [
    "data_out = np.hstack((np.ones((train_data.shape[0], 1)), train_data))\n",
    "test_out = np.hstack((np.ones((test_data.shape[0], 1)), test_data))\n",
    "\n",
    "init_weights = np.zeros((data_out.shape[1], 1))\n",
    "weights_1 = logistic_regpression_regularized(data_out, train_label, init_weights, 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "9843b52d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "res = sigmoid(np.dot(test_out, weights_1))\n",
    "preds = []\n",
    "for i in res:\n",
    "    if i >= 0.5:\n",
    "        preds.append(1)\n",
    "    else:\n",
    "        preds.append(0)\n",
    "\n",
    "preds = np.array(preds)\n",
    "conf_matrix = confusion_matrix(y_true = test_label.reshape(-1,), y_pred = preds)\n",
    "df_cmt = pd.DataFrame(conf_matrix, index = ['Healthy', 'Covid'], columns = ['Healthy', 'Covid'])\n",
    "plt.figure(figsize = (10,7))\n",
    "sn.heatmap(df_cmt, annot=True, cmap=\"YlGnBu\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2a176dd7",
   "metadata": {},
   "source": [
    "### Check confusion matrix with sklearn.linear_model.LogisticRegression and plot the ROC curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "a360c018",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "\n",
    "lr = LogisticRegression(penalty='l2', C=10)\n",
    "lr.fit(train_data, train_label.ravel())\n",
    "\n",
    "preds = lr.predict(test_data)\n",
    "conf_matrix = confusion_matrix(y_true = test_label.reshape(-1,), y_pred = preds)\n",
    "df_cmt = pd.DataFrame(conf_matrix, index = ['Healthy', 'Covid'], columns = ['Healthy', 'Covid'])\n",
    "plt.figure(figsize = (10,7))\n",
    "sn.heatmap(df_cmt, annot=True, cmap=\"YlGnBu\")\n",
    "\n",
    "probas = lr.predict_proba(test_data[0:n_test])[:, 1]\n",
    "fpr, tpr, thresholds = roc_curve(test_label[0:n_test], probas)\n",
    "roc_auc = auc(fpr, tpr)\n",
    "\n",
    "plt.figure()\n",
    "plt.clf()\n",
    "plt.plot(fpr, tpr, label='ROC curve (area = %0.2f)' % roc_auc)\n",
    "plt.plot([0, 1], [0, 1], 'k--')\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.ylim([0.0, 1.0])\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.title('ROC Curve for Logistic regression with L2 regularization')\n",
    "plt.legend(loc=\"upper left\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d983f4d6",
   "metadata": {},
   "source": [
    "### ROC Curve for Decision tree"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "57bd30c2",
   "metadata": {},
   "source": [
    "### Plot confusion matrix with the result from decisionTree.py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "f070998a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "tree_out = pd.read_csv(\"test_outputs.csv\", header=None).to_numpy()\n",
    "conf_matrix = confusion_matrix(y_true = test_label.reshape(-1,), y_pred = tree_out)\n",
    "df_cmt = pd.DataFrame(conf_matrix, index = ['Healthy', 'Covid'], columns = ['Healthy', 'Covid'])\n",
    "plt.figure(figsize = (10,7))\n",
    "sn.heatmap(df_cmt, annot=True, cmap=\"YlGnBu\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "01c36919",
   "metadata": {},
   "source": [
    "### Check confusion matrix with sklearn.tree.DecisionTreeClassifier and plot the ROC curve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "1d8d2b36",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.tree import DecisionTreeClassifier\n",
    "\n",
    "model = DecisionTreeClassifier()\n",
    "model.fit(train_data, train_label.ravel())\n",
    "\n",
    "preds = model.predict(test_data)\n",
    "conf_matrix = confusion_matrix(y_true = test_label.reshape(-1,), y_pred = preds)\n",
    "df_cmt = pd.DataFrame(conf_matrix, index = ['Healthy', 'Covid'], columns = ['Healthy', 'Covid'])\n",
    "plt.figure(figsize = (10,7))\n",
    "sn.heatmap(df_cmt, annot=True, cmap=\"YlGnBu\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "efc0c5db",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "probas = model.predict_proba(test_data)[:, 1]\n",
    "\n",
    "fpr, tpr, thresholds = roc_curve(test_label, probas)\n",
    "roc_auc = auc(fpr, tpr)\n",
    "\n",
    "plt.clf()\n",
    "plt.plot(fpr, tpr, label='ROC curve (area = %0.2f)' % roc_auc)\n",
    "plt.plot([0, 1], [0, 1], 'k--')\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.ylim([0.0, 1.0])\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.title('ROC Curve for Decision tree')\n",
    "plt.legend(loc=\"upper left\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f50e2a93",
   "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
}