Capstone Project 3.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Movie reviews using Recurrent Nerual Networks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import libraries\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "\n",
    "from tensorflow.keras.datasets import imdb\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.layers import LSTM, Flatten, Dropout\n",
    "from tensorflow.keras.layers import Embedding\n",
    "from tensorflow.keras.preprocessing import sequence\n",
    "\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load IMDB from Keras dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Vocabulary size is 2000\n",
    "num_words = 2000\n",
    "(X_train, y_train), (X_test, y_test) = imdb.load_data(path=\"imdb.npz\",\n",
    "                                                      num_words=num_words,\n",
    "                                                      skip_top=0,\n",
    "                                                      maxlen=None,\n",
    "                                                      seed=113,\n",
    "                                                      start_char=1,\n",
    "                                                      oov_char=2,\n",
    "                                                      index_from=3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Pad the sequence to the same length"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "max_review_length = 500\n",
    "X_train = sequence.pad_sequences(X_train, maxlen=max_review_length)\n",
    "X_test = sequence.pad_sequences(X_test, maxlen=max_review_length)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create and compile model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_1 (Embedding)      (None, 500, 32)           64000     \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 500, 32)           0         \n",
      "_________________________________________________________________\n",
      "lstm_1 (LSTM)                (None, 32)                8320      \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 256)               8448      \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 1)                 257       \n",
      "=================================================================\n",
      "Total params: 81,025\n",
      "Trainable params: 81,025\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "embedding_vector_length = 32\n",
    "model = Sequential()\n",
    "model.add(Embedding(input_dim=num_words, output_dim=embedding_vector_length, input_length=max_review_length))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(LSTM(32))\n",
    "model.add(Dense(units=256, activation='relu'))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(Dense(units=1, activation='sigmoid'))\n",
    "model.summary()\n",
    "model.compile(loss='binary_crossentropy',\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Fit The Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 20000 samples, validate on 5000 samples\n",
      "Epoch 1/10\n",
      "20000/20000 - 128s - loss: 0.4631 - acc: 0.7692 - val_loss: 0.3817 - val_acc: 0.8298\n",
      "Epoch 2/10\n",
      "20000/20000 - 130s - loss: 0.3229 - acc: 0.8648 - val_loss: 0.3861 - val_acc: 0.8222\n",
      "Epoch 3/10\n",
      "20000/20000 - 120s - loss: 0.2810 - acc: 0.8864 - val_loss: 0.3789 - val_acc: 0.8596\n",
      "Epoch 4/10\n",
      "20000/20000 - 127s - loss: 0.2610 - acc: 0.8979 - val_loss: 0.3030 - val_acc: 0.8762\n",
      "Epoch 5/10\n",
      "20000/20000 - 123s - loss: 0.2406 - acc: 0.9028 - val_loss: 0.3200 - val_acc: 0.8720\n",
      "Epoch 6/10\n",
      "20000/20000 - 133s - loss: 0.2299 - acc: 0.9090 - val_loss: 0.3399 - val_acc: 0.8698\n",
      "Epoch 7/10\n",
      "20000/20000 - 137s - loss: 0.2156 - acc: 0.9149 - val_loss: 0.3610 - val_acc: 0.8602\n",
      "Epoch 8/10\n",
      "20000/20000 - 137s - loss: 0.2004 - acc: 0.9212 - val_loss: 0.3362 - val_acc: 0.8654\n",
      "Epoch 9/10\n",
      "20000/20000 - 132s - loss: 0.1974 - acc: 0.9237 - val_loss: 0.3308 - val_acc: 0.8676\n",
      "Epoch 10/10\n",
      "20000/20000 - 121s - loss: 0.1869 - acc: 0.9263 - val_loss: 0.4439 - val_acc: 0.8642\n"
     ]
    }
   ],
   "source": [
    "train_history = model.fit(X_train, y_train, batch_size=32,\n",
    "                          epochs=10, verbose=2,\n",
    "                          validation_split=0.2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Show train and validation history: loss and accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "def show_train_history(train_history,train,validation):\n",
    "    plt.plot(train_history.history[train])\n",
    "    plt.plot(train_history.history[validation])\n",
    "    plt.title('Train History')\n",
    "    plt.ylabel(train)\n",
    "    plt.xlabel('Epoch')\n",
    "    plt.legend(['train', 'validation'], loc='upper left')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "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": [
    "show_train_history(train_history,'acc','val_acc')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "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": [
    "show_train_history(train_history,'loss','val_loss')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate model by test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "25000/25000 [==============================] - 32s 1ms/sample - loss: 0.4368 - acc: 0.8641\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.86412"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scores = model.evaluate(X_test, y_test, verbose=1)\n",
    "scores[1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Predict test data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "predict=model.predict_classes(X_test)\n",
    "predict_classes=predict.reshape(len(X_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_original_text(i):\n",
    "    word_to_id = imdb.get_word_index()\n",
    "    word_to_id = {k:(v+3) for k,v in word_to_id.items()}\n",
    "    word_to_id[\"<PAD>\"] = 0\n",
    "    word_to_id[\"<START>\"] = 1\n",
    "    word_to_id[\"<UNK>\"] = 2\n",
    "\n",
    "    id_to_word = {value:key for key,value in word_to_id.items()}\n",
    "    return ' '.join(id_to_word[id] for id in X_test[i])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "SentimentDict={1:'positive', 0:'negative'}\n",
    "def display_test_sentiment(i):\n",
    "    print(get_original_text(i))\n",
    "    print('label: ', SentimentDict[y_test[i]], ', prediction: ', SentimentDict[predict_classes[i]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <START> i'm absolutely <UNK> this movie isn't being <UNK> all who love this movie should <UNK> disney and <UNK> the <UNK> for it <UNK> eventually have to <UNK> it then i'd buy <UNK> for everybody i know everything and everybody in this movie did a good job and i haven't <UNK> out why disney hasn't put this movie on dvd or on vhs in <UNK> <UNK> at least i haven't seen any <UNK> this is a <UNK> good movie and should be seen by all the kids in the new <UNK> don't get to see it and i think they should it should at least be put back on the channel this movie doesn't deserve a cheap <UNK> it deserves the real thing i'm them now this movie will be on dvd\n",
      "label:  positive , prediction:  positive\n"
     ]
    }
   ],
   "source": [
    "display_test_sentiment(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <START> i saw this movie on <UNK> with my younger brother a long time ago despite the fact i was a young boy who's <UNK> <UNK> i didn't have the <UNK> <UNK> <UNK> at the time it was <UNK> the year of the <UNK> the movie but i could follow the plot and understand what's going on it wasn't my <UNK> i saw the animated movie intended for girls my father rented the video to show to my other younger sister br br a <UNK> later and i rented the video for 50 <UNK> to watch again with the <UNK> <UNK> turned on my memories of this movie was utterly <UNK> by none other than a <UNK> script i have seen plenty of poorly written movies like cool as ice and jason goes to hell the final <UNK> but i have never seen or heard the <UNK> this bad only with enough <UNK> to make your head <UNK> from laughing in <UNK> and screaming from the pain of <UNK> the torture of sitting through this movie despite good plot and intriguing story <UNK> the script has to be one of the worst ever written for the screen bar none the <UNK> <UNK> r <UNK> should never be working as a <UNK> professional or otherwise i can not believe they would even allow the terrible script to <UNK> a movie like this in the first place did the japanese producers read the script in broken english or <UNK> before they know what they were into even crap like g i joe the movie and my little <UNK> the movie have redeeming values compared to this <UNK> br br if you're a big fan of <UNK> animation or just taking a <UNK> trip <UNK> of <UNK> and the star <UNK> it does not matter whether you were <UNK> or <UNK> by the <UNK> animated movie with an <UNK> written script or you have not seen the movie stay away from this movie the movie should be <UNK> with the to learn how not to write a bad script\n",
      "label:  negative , prediction:  negative\n"
     ]
    }
   ],
   "source": [
    "display_test_sentiment(12000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}