4.TEXT_CLASSIFICATION/4.1.6 RNN Classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 모듈 불러오기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "pycharm": {
     "is_executing": false
    }
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "from tensorflow.keras import layers\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import os\n",
    "import json\n",
    "\n",
    "from tqdm import tqdm"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 시각화 함수"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "pycharm": {
     "is_executing": false
    }
   },
   "outputs": [],
   "source": [
    "def plot_graphs(history, string):\n",
    "    plt.plot(history.history[string])\n",
    "    plt.plot(history.history['val_'+string], '')\n",
    "    plt.xlabel(\"Epochs\")\n",
    "    plt.ylabel(string)\n",
    "    plt.legend([string, 'val_'+string])\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 학습 데이터 경로 정의 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "pycharm": {
     "is_executing": false
    }
   },
   "outputs": [],
   "source": [
    "DATA_IN_PATH = './data_in/'\n",
    "DATA_OUT_PATH = './data_out/'\n",
    "TRAIN_INPUT_DATA = 'train_input.npy'\n",
    "TRAIN_LABEL_DATA = 'train_label.npy'\n",
    "DATA_CONFIGS = 'data_configs.json'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 랜덤 시드 고정"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "SEED_NUM = 1234\n",
    "tf.random.set_seed(SEED_NUM)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 파일 로드"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_input = np.load(open(DATA_IN_PATH + TRAIN_INPUT_DATA, 'rb'))\n",
    "train_label = np.load(open(DATA_IN_PATH + TRAIN_LABEL_DATA, 'rb'))\n",
    "prepro_configs = json.load(open(DATA_IN_PATH + DATA_CONFIGS, 'r'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 모델 하이퍼파라메터 정의"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "pycharm": {
     "is_executing": false
    }
   },
   "outputs": [],
   "source": [
    "model_name = 'rnn_classifier_en'\n",
    "BATCH_SIZE = 128\n",
    "NUM_EPOCHS = 5\n",
    "VALID_SPLIT = 0.1\n",
    "MAX_LEN = train_input.shape[1]\n",
    "\n",
    "kargs = {'model_name': model_name,\n",
    "        'vocab_size': prepro_configs['vocab_size'],\n",
    "        'embedding_dimension': 100,\n",
    "        'dropout_rate': 0.2,\n",
    "        'lstm_dimension': 150,\n",
    "        'dense_dimension': 150,\n",
    "        'output_dimension':1}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 모델 선언 및 컴파일"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "pycharm": {
     "is_executing": false
    }
   },
   "outputs": [],
   "source": [
    "class RNNClassifier(tf.keras.Model):\n",
    "    def __init__(self, **kargs):\n",
    "        super(RNNClassifier, self).__init__(name=kargs['model_name'])\n",
    "        self.embedding = layers.Embedding(input_dim=kargs['vocab_size'],\n",
    "                                     output_dim=kargs['embedding_dimension'])\n",
    "        self.lstm_1_layer = tf.keras.layers.LSTM(kargs['lstm_dimension'], return_sequences=True)\n",
    "        self.lstm_2_layer = tf.keras.layers.LSTM(kargs['lstm_dimension'])\n",
    "        self.dropout = layers.Dropout(kargs['dropout_rate'])\n",
    "        self.fc1 = layers.Dense(units=kargs['dense_dimension'],\n",
    "                           activation=tf.keras.activations.tanh)\n",
    "        self.fc2 = layers.Dense(units=kargs['output_dimension'],\n",
    "                           activation=tf.keras.activations.sigmoid)\n",
    "    \n",
    "    def call(self, x):\n",
    "        x = self.embedding(x)\n",
    "        x = self.dropout(x)\n",
    "        x = self.lstm_1_layer(x)\n",
    "        x = self.lstm_2_layer(x)\n",
    "        x = self.dropout(x)\n",
    "        x = self.fc1(x)\n",
    "        x = self.dropout(x)\n",
    "        x = self.fc2(x)\n",
    "        \n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = RNNClassifier(**kargs)\n",
    "model.compile(optimizer=tf.keras.optimizers.Adam(1e-4),\n",
    "              loss=tf.keras.losses.BinaryCrossentropy(),\n",
    "              metrics=[tf.keras.metrics.BinaryAccuracy(name='accuracy')])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Callback 선언"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "pycharm": {
     "is_executing": false
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "./data_out/rnn_classifier_en -- Folder already exists \n",
      "\n"
     ]
    }
   ],
   "source": [
    "# overfitting을 막기 위한 ealrystop 추가\n",
    "earlystop_callback = EarlyStopping(monitor='val_accuracy', min_delta=0.0001, patience=1)\n",
    "# min_delta: the threshold that triggers the termination (acc should at least improve 0.0001)\n",
    "# patience: no improvment epochs (patience = 1, 1번 이상 상승이 없으면 종료)\n",
    "\n",
    "checkpoint_path = DATA_OUT_PATH + model_name + '/weights.h5'\n",
    "checkpoint_dir = os.path.dirname(checkpoint_path)\n",
    "\n",
    "# Create path if exists\n",
    "if os.path.exists(checkpoint_dir):\n",
    "    print(\"{} -- Folder already exists \\n\".format(checkpoint_dir))\n",
    "else:\n",
    "    os.makedirs(checkpoint_dir, exist_ok=True)\n",
    "    print(\"{} -- Folder create complete \\n\".format(checkpoint_dir))\n",
    "    \n",
    "\n",
    "cp_callback = ModelCheckpoint(\n",
    "    checkpoint_path, monitor='val_accuracy', verbose=1, save_best_only=True, save_weights_only=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 모델 학습"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 22500 samples, validate on 2500 samples\n",
      "Epoch 1/5\n",
      "22400/22500 [============================>.] - ETA: 0s - loss: 0.6931 - accuracy: 0.5059\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.51040, saving model to ./data_out/rnn_classifier_en/weights.h5\n",
      "22500/22500 [==============================] - 162s 7ms/sample - loss: 0.6931 - accuracy: 0.5060 - val_loss: 0.6927 - val_accuracy: 0.5104\n",
      "Epoch 2/5\n",
      "22400/22500 [============================>.] - ETA: 0s - loss: 0.6894 - accuracy: 0.5236\n",
      "Epoch 00002: val_accuracy improved from 0.51040 to 0.54960, saving model to ./data_out/rnn_classifier_en/weights.h5\n",
      "22500/22500 [==============================] - 159s 7ms/sample - loss: 0.6893 - accuracy: 0.5240 - val_loss: 0.6825 - val_accuracy: 0.5496\n",
      "Epoch 3/5\n",
      "22400/22500 [============================>.] - ETA: 0s - loss: 0.4373 - accuracy: 0.7844\n",
      "Epoch 00003: val_accuracy improved from 0.54960 to 0.87920, saving model to ./data_out/rnn_classifier_en/weights.h5\n",
      "22500/22500 [==============================] - 176s 8ms/sample - loss: 0.4363 - accuracy: 0.7850 - val_loss: 0.2902 - val_accuracy: 0.8792\n",
      "Epoch 4/5\n",
      "22400/22500 [============================>.] - ETA: 0s - loss: 0.2168 - accuracy: 0.9175\n",
      "Epoch 00004: val_accuracy improved from 0.87920 to 0.89080, saving model to ./data_out/rnn_classifier_en/weights.h5\n",
      "22500/22500 [==============================] - 168s 7ms/sample - loss: 0.2165 - accuracy: 0.9176 - val_loss: 0.2722 - val_accuracy: 0.8908\n",
      "Epoch 5/5\n",
      "22400/22500 [============================>.] - ETA: 0s - loss: 0.1397 - accuracy: 0.9528\n",
      "Epoch 00005: val_accuracy did not improve from 0.89080\n",
      "22500/22500 [==============================] - 174s 8ms/sample - loss: 0.1396 - accuracy: 0.9528 - val_loss: 0.3151 - val_accuracy: 0.8908\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(train_input, train_label, batch_size=BATCH_SIZE, epochs=NUM_EPOCHS,\n",
    "                    validation_split=VALID_SPLIT, callbacks=[earlystop_callback, cp_callback])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 결과 플롯"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "pycharm": {
     "is_executing": false
    }
   },
   "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": [
    "plot_graphs(history, 'accuracy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "pycharm": {
     "is_executing": false
    }
   },
   "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": [
    "plot_graphs(history, 'loss')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 테스트 데이터 불러오기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "pycharm": {
     "is_executing": false
    }
   },
   "outputs": [],
   "source": [
    "DATA_OUT_PATH = './data_out/'\n",
    "TEST_INPUT_DATA = 'test_input.npy'\n",
    "TEST_ID_DATA = 'test_id.npy'\n",
    "\n",
    "\n",
    "test_input = np.load(open(DATA_IN_PATH + TEST_INPUT_DATA, 'rb'))\n",
    "test_input = pad_sequences(test_input, maxlen=test_input.shape[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 베스트 모델 불러오기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "SAVE_FILE_NM = 'weights.h5'\n",
    "\n",
    "model.load_weights(os.path.join(DATA_OUT_PATH, model_name, SAVE_FILE_NM))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 테스트 데이터 예측하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions = model.predict(test_input, batch_size=BATCH_SIZE)\n",
    "predictions = predictions.squeeze(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "pycharm": {
     "is_executing": false
    }
   },
   "outputs": [],
   "source": [
    "test_id = np.load(open(DATA_IN_PATH + TEST_ID_DATA, 'rb'), allow_pickle=True)\n",
    "\n",
    "if not os.path.exists(DATA_OUT_PATH):\n",
    "    os.makedirs(DATA_OUT_PATH)\n",
    "\n",
    "output = pd.DataFrame(data={\"id\": list(test_id), \"sentiment\":list(predictions)})\n",
    "output.to_csv(DATA_OUT_PATH + 'movie_review_result_rnn.csv', index=False, quoting=3)"
   ]
  }
 ],
 "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"
  },
  "pycharm": {
   "stem_cell": {
    "cell_type": "raw",
    "metadata": {
     "collapsed": false
    },
    "source": []
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}