NLP_basic/imdb_basic.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "#IMDB는 Movie Review Classification을 위한 Dataset이며, 스탠포드 연구자들이 2011년도에 \n",
    "#총 50000개의 데이터를 반은 트레이닝 반은 테스트로 사용하여 89%의 성능을 얻어냈다.\n",
    " \n",
    "import numpy\n",
    "from matplotlib import pyplot\n",
    "from keras.datasets import imdb\n",
    "from keras.models import Sequential\n",
    "from keras.layers.embeddings import Embedding\n",
    "from keras.preprocessing import sequence\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "from keras.layers import Dense, Dropout, Activation, Flatten\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.layers.convolutional import Convolution1D\n",
    "from keras.layers.convolutional import MaxPooling1D\n",
    "from matplotlib import pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# MLP for the IMDB problem\n",
    "# fix random seed for reproducibility\n",
    "seed = 7\n",
    "numpy.random.seed(seed)\n",
    "# load the dataset but only keep the top n words, zero the rest\n",
    "top_words = 5000\n",
    "test_split = 0.33\n",
    "(x_train, y_train), (x_test, y_test) = imdb.load_data(num_words=top_words)\n",
    "max_words = 500\n",
    "x_train = sequence.pad_sequences(x_train, maxlen=max_words)\n",
    "x_test = sequence.pad_sequences(x_test, maxlen=max_words)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\kimta\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:4: UserWarning: Update your `Conv1D` call to the Keras 2 API: `Conv1D(filters=32, kernel_size=3, padding=\"same\")`\n",
      "  after removing the cwd from sys.path.\n",
      "C:\\Users\\kimta\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:7: UserWarning: Update your `MaxPooling1D` call to the Keras 2 API: `MaxPooling1D(pool_size=2)`\n",
      "  import sys\n",
      "C:\\Users\\kimta\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:8: UserWarning: Update your `Conv1D` call to the Keras 2 API: `Conv1D(filters=32, kernel_size=3, padding=\"same\")`\n",
      "  \n",
      "C:\\Users\\kimta\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:11: UserWarning: Update your `MaxPooling1D` call to the Keras 2 API: `MaxPooling1D(pool_size=2)`\n",
      "  # This is added back by InteractiveShellApp.init_path()\n",
      "C:\\Users\\kimta\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:12: UserWarning: Update your `Conv1D` call to the Keras 2 API: `Conv1D(filters=32, kernel_size=3, padding=\"same\")`\n",
      "  if sys.path[0] == '':\n",
      "C:\\Users\\kimta\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:15: UserWarning: Update your `MaxPooling1D` call to the Keras 2 API: `MaxPooling1D(pool_size=2)`\n",
      "  from ipykernel import kernelapp as app\n",
      "C:\\Users\\kimta\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:29: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_1 (Embedding)      (None, 500, 32)           160000    \n",
      "_________________________________________________________________\n",
      "conv1d_1 (Conv1D)            (None, 500, 32)           3104      \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 500, 32)           0         \n",
      "_________________________________________________________________\n",
      "max_pooling1d_1 (MaxPooling1 (None, 250, 32)           0         \n",
      "_________________________________________________________________\n",
      "conv1d_2 (Conv1D)            (None, 250, 32)           3104      \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 250, 32)           0         \n",
      "_________________________________________________________________\n",
      "max_pooling1d_2 (MaxPooling1 (None, 125, 32)           0         \n",
      "_________________________________________________________________\n",
      "conv1d_3 (Conv1D)            (None, 125, 32)           3104      \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 125, 32)           0         \n",
      "_________________________________________________________________\n",
      "activation_3 (Activation)    (None, 125, 32)           0         \n",
      "_________________________________________________________________\n",
      "max_pooling1d_3 (MaxPooling1 (None, 62, 32)            0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 1984)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 250)               496250    \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 250)               0         \n",
      "_________________________________________________________________\n",
      "activation_4 (Activation)    (None, 250)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 250)               62750     \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 250)               0         \n",
      "_________________________________________________________________\n",
      "activation_5 (Activation)    (None, 250)               0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 1)                 251       \n",
      "_________________________________________________________________\n",
      "activation_6 (Activation)    (None, 1)                 0         \n",
      "=================================================================\n",
      "Total params: 728,563\n",
      "Trainable params: 728,563\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 25000 samples, validate on 25000 samples\n",
      "Epoch 1/2\n",
      "25000/25000 [==============================] - 6s 225us/step - loss: 0.4425 - acc: 0.7584 - val_loss: 0.3129 - val_acc: 0.8636\n",
      "Epoch 2/2\n",
      "25000/25000 [==============================] - 3s 140us/step - loss: 0.2130 - acc: 0.9168 - val_loss: 0.2604 - val_acc: 0.8939\n",
      "Accuracy: 89.39%\n"
     ]
    }
   ],
   "source": [
    "# create the model\n",
    "model = Sequential()\n",
    "model.add(Embedding(top_words, 32, input_length=max_words))\n",
    "model.add(Convolution1D(nb_filter=32, filter_length=3, border_mode='same'))\n",
    "#model.add(Dropout(0.25))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling1D(pool_length=2))\n",
    "model.add(Convolution1D(nb_filter=32, filter_length=3, border_mode='same'))\n",
    "#model.add(Dropout(0.25))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling1D(pool_length=2))\n",
    "model.add(Convolution1D(nb_filter=32, filter_length=3, border_mode='same'))\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling1D(pool_length=2))\n",
    "model.add(Flatten())\n",
    "model.add(Dense(250))\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dense(250))\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dense(1))\n",
    "model.add(Activation('sigmoid'))\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "print(model.summary())\n",
    "\n",
    "# Fit the model\n",
    "hist = model.fit(x_train, y_train, validation_data=(x_test, y_test), nb_epoch=2, batch_size=64, verbose=1)\n",
    "\n",
    "# Final evaluation of the model\n",
    "scores = model.evaluate(x_test, y_test, verbose=0)\n",
    "print(\"Accuracy: %.2f%%\" % (scores[1]*100))\n",
    " \n",
    " \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#모델 시각\n",
    "fig, loss_ax = plt.subplots()\n",
    " \n",
    "acc_ax = loss_ax.twinx()\n",
    " \n",
    "loss_ax.plot(hist.history['loss'], 'y', label='train loss')\n",
    "loss_ax.plot(hist.history['val_loss'], 'r', label='val loss')\n",
    " \n",
    "acc_ax.plot(hist.history['acc'], 'b', label='train acc')\n",
    "acc_ax.plot(hist.history['val_acc'], 'g', label='val acc')\n",
    " \n",
    "loss_ax.set_xlabel('epoch')\n",
    "loss_ax.set_ylabel('loss')\n",
    "acc_ax. set_ylabel('accuracy')\n",
    " \n",
    "loss_ax.legend(loc='upper left')\n",
    "acc_ax.legend(loc='lower left')\n",
    " \n",
    "plt.show()\n",
    " \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}