[E-06]MovieReview.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "3593c71e",
   "metadata": {},
   "source": [
    "# 네이버 영화리뷰 감성분석 도전하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "721e335c",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import os"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c3b34bd4",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e4d5cd9e",
   "metadata": {},
   "source": [
    "# 1. 데이터 준비와 확인"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "7f6b844a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(150000, 3)\n"
     ]
    },
    {
     "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>id</th>\n",
       "      <th>document</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>9976970</td>\n",
       "      <td>아 더빙.. 진짜 짜증나네요 목소리</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3819312</td>\n",
       "      <td>흠...포스터보고 초딩영화줄....오버연기조차 가볍지 않구나</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>10265843</td>\n",
       "      <td>너무재밓었다그래서보는것을추천한다</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>9045019</td>\n",
       "      <td>교도소 이야기구먼 ..솔직히 재미는 없다..평점 조정</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6483659</td>\n",
       "      <td>사이몬페그의 익살스런 연기가 돋보였던 영화!스파이더맨에서 늙어보이기만 했던 커스틴 ...</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         id                                           document  label\n",
       "0   9976970                                아 더빙.. 진짜 짜증나네요 목소리      0\n",
       "1   3819312                  흠...포스터보고 초딩영화줄....오버연기조차 가볍지 않구나      1\n",
       "2  10265843                                  너무재밓었다그래서보는것을추천한다      0\n",
       "3   9045019                      교도소 이야기구먼 ..솔직히 재미는 없다..평점 조정      0\n",
       "4   6483659  사이몬페그의 익살스런 연기가 돋보였던 영화!스파이더맨에서 늙어보이기만 했던 커스틴 ...      1"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_data = pd.read_table('/aiffel/data/ratings_train.txt')\n",
    "test_data = pd.read_table('/aiffel/data/ratings_test.txt')\n",
    "\n",
    "print(train_data.shape)\n",
    "train_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "b56a3b01",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(50000, 3)\n"
     ]
    },
    {
     "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>id</th>\n",
       "      <th>document</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6270596</td>\n",
       "      <td>굳 ㅋ</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>9274899</td>\n",
       "      <td>GDNTOPCLASSINTHECLUB</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>8544678</td>\n",
       "      <td>뭐야 이 평점들은.... 나쁘진 않지만 10점 짜리는 더더욱 아니잖아</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>6825595</td>\n",
       "      <td>지루하지는 않은데 완전 막장임... 돈주고 보기에는....</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6723715</td>\n",
       "      <td>3D만 아니었어도 별 다섯 개 줬을텐데.. 왜 3D로 나와서 제 심기를 불편하게 하죠??</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        id                                           document  label\n",
       "0  6270596                                                굳 ㅋ      1\n",
       "1  9274899                               GDNTOPCLASSINTHECLUB      0\n",
       "2  8544678             뭐야 이 평점들은.... 나쁘진 않지만 10점 짜리는 더더욱 아니잖아      0\n",
       "3  6825595                   지루하지는 않은데 완전 막장임... 돈주고 보기에는....      0\n",
       "4  6723715  3D만 아니었어도 별 다섯 개 줬을텐데.. 왜 3D로 나와서 제 심기를 불편하게 하죠??      0"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(test_data.shape)\n",
    "test_data.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4aa05f46",
   "metadata": {},
   "source": [
    "# 2. 데이터 구성\n",
    "\n",
    "1. docoment 컬럼 기준으로 데이터 중복 제거\n",
    "2. NaN 하나라도 컬럼에 있으면 결측치 제거\n",
    "3. Mecab으로 형태소 추출 -> 토큰화\n",
    "4. 안쓰는 불용어(stopwords)는 빼기\n",
    "5. 토큰들(단어) 각각 개수를 세봤을 때, 많은 것 순으로 정렬한 뒤 {10000-4}개 단어들만 가져와서\n",
    "    - 별도 `'<PAD>', '<BOS>', '<UNK>', '<UNUSED>'` 추가\n",
    "6. 단어 딕셔너리 만들기; word_to_index, index_to_word"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 675,
   "id": "0beb7482",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "This is sample sentence. \"안녕하세요~~~~~\" 주인공이 말했다 234246\n"
     ]
    }
   ],
   "source": [
    "import re\n",
    "\n",
    "def preprocess_sentence(sentence):\n",
    "#     sentence = re.sub(r\"[!@#$%^&*()~!?\\'\\\".,]+\", \" \", sentence)\n",
    "#     sentence = re.sub(r\"[^ 가-힣A-Za-z0-9]+\", \" \", sentence)\n",
    "#     sentence = re.sub(r'[\" \"]+', \" \", sentence)\n",
    "#     sentence = sentence.strip()\n",
    "    return sentence\n",
    "\n",
    "\n",
    "print(preprocess_sentence(\"This is sample sentence. \\\"안녕하세요~~~~~\\\" 주인공이 말했다 234246\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 676,
   "id": "947622b5",
   "metadata": {},
   "outputs": [],
   "source": [
    "from konlpy.tag import Mecab\n",
    "from collections import Counter\n",
    "\n",
    "tokenizer = Mecab()\n",
    "stopwords = ['의','가','이','은','들','는','좀','잘','걍','과','도','를','으로','자','에','와','한','하다', '음']\n",
    "\n",
    "def load_data(train_data, test_data, num_words=10000):\n",
    "    train_data.drop_duplicates(subset=['document'], inplace=True)\n",
    "    train_data = train_data.dropna(how = 'any') \n",
    "    test_data.drop_duplicates(subset=['document'], inplace=True)\n",
    "    test_data = test_data.dropna(how = 'any') \n",
    "    \n",
    "    X_train = []\n",
    "    for sentence in train_data['document']:\n",
    "        s = preprocess_sentence(sentence)\n",
    "        temp_X = tokenizer.morphs(s) # 토큰화\n",
    "        temp_X = [word for word in temp_X if not word in stopwords] # 불용어 제거\n",
    "        X_train.append(temp_X)\n",
    "\n",
    "    X_test = []\n",
    "    for sentence in test_data['document']:\n",
    "        s = preprocess_sentence(sentence)\n",
    "        temp_X = tokenizer.morphs(s) # 토큰화\n",
    "        temp_X = [word for word in temp_X if not word in stopwords] # 불용어 제거\n",
    "        X_test.append(temp_X)\n",
    "    \n",
    "    words = np.concatenate(X_train).tolist()\n",
    "    counter = Counter(words)\n",
    "    counter = counter.most_common(num_words-4) # 많은 것 순으로 정렬한 뒤, {num_words-4}개 단어들만 반환\n",
    "    vocab = ['<PAD>', '<BOS>', '<UNK>', '<UNUSED>'] + [key for key, _ in counter]\n",
    "    word_to_index = {word:index for index, word in enumerate(vocab)}\n",
    "        \n",
    "    def get_indexlist(wordlist):\n",
    "        return [word_to_index[word] if word in word_to_index else word_to_index['<UNK>'] for word in wordlist]\n",
    "        \n",
    "    X_train = list(map(get_indexlist, X_train)) #\n",
    "    X_test = list(map(get_indexlist, X_test))\n",
    "        \n",
    "    return X_train, np.array(list(train_data['label'])), X_test, np.array(list(test_data['label'])), word_to_index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 677,
   "id": "ede9c6fd",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, y_train, X_test, y_test, word_to_index = load_data(train_data, test_data, 10000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 678,
   "id": "1354ee59",
   "metadata": {},
   "outputs": [],
   "source": [
    "index_to_word = {index:word for word, index in word_to_index.items()}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 679,
   "id": "95ac9f86",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "146182"
      ]
     },
     "execution_count": 679,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(X_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c387688f",
   "metadata": {},
   "source": [
    "### 데이터 확인"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 680,
   "id": "6567b3c3",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 문장 1개를 활용할 딕셔너리와 함께 주면, 단어 인덱스 리스트 벡터로 변환해 주는 함수입니다. \n",
    "# 단, 모든 문장은 <BOS>로 시작하는 것으로 합니다. \n",
    "def get_encoded_sentence(sentence, word_to_index):\n",
    "    return [word_to_index['<BOS>']]+[word_to_index[word] if word in word_to_index else word_to_index['<UNK>'] for word in sentence.split()]\n",
    "\n",
    "# 여러 개의 문장 리스트를 한꺼번에 단어 인덱스 리스트 벡터로 encode해 주는 함수입니다. \n",
    "def get_encoded_sentences(sentences, word_to_index):\n",
    "    return [get_encoded_sentence(sentence, word_to_index) for sentence in sentences]\n",
    "\n",
    "# 숫자 벡터로 encode된 문장을 원래대로 decode하는 함수입니다. \n",
    "def get_decoded_sentence(encoded_sentence, index_to_word):\n",
    "    return ' '.join(index_to_word[index] if index in index_to_word else '<UNK>' for index in encoded_sentence[1:])  #[1:]를 통해 <BOS>를 제외\n",
    "\n",
    "# 여러 개의 숫자 벡터로 encode된 문장을 한꺼번에 원래대로 decode하는 함수입니다. \n",
    "def get_decoded_sentences(encoded_sentences, index_to_word):\n",
    "    return [get_decoded_sentence(encoded_sentence, index_to_word) for encoded_sentence in encoded_sentences]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 730,
   "id": "84cf2088",
   "metadata": {},
   "outputs": [],
   "source": [
    "# get_decoded_sentences(X_train, index_to_word)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "537c2914",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "# 3. 모델 구성을 위한 데이터 분석 및 가공\n",
    "\n",
    "- 데이터셋 내 문장 길이 분포\n",
    "- 적절한 최대 문장 길이 지정\n",
    "- keras.preprocessing.sequence.pad_sequences 을 활용한 패딩 추가"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "020b8054",
   "metadata": {},
   "source": [
    "### 데이터셋 내 문장 길이 분포"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 682,
   "id": "cddbee8c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "문장길이 평균 :  15.905482264166398\n",
      "문장길이 최대 :  116\n",
      "문장길이 표준편차 :  12.809775372249128\n"
     ]
    }
   ],
   "source": [
    "total_data_text = list(X_train) + list(X_test)\n",
    "\n",
    "num_tokens = [len(tokens) for tokens in total_data_text]\n",
    "num_tokens = np.array(num_tokens)\n",
    "\n",
    "print('문장길이 평균 : ', np.mean(num_tokens))\n",
    "print('문장길이 최대 : ', np.max(num_tokens))\n",
    "print('문장길이 표준편차 : ', np.std(num_tokens))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 683,
   "id": "69201534",
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "427dc1ae",
   "metadata": {},
   "source": [
    "### 적절한 최대 문장 길이 지정"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 684,
   "id": "19fe0dfa",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "pad_sequences maxlen :  41\n",
      "전체 문장의 0.9347083787671688%가 maxlen 설정값 이내에 포함됩니다. \n"
     ]
    }
   ],
   "source": [
    "max_tokens = np.mean(num_tokens) + 2 * np.std(num_tokens)\n",
    "maxlen = int(max_tokens)\n",
    "\n",
    "print('pad_sequences maxlen : ', maxlen)\n",
    "print('전체 문장의 {}%가 maxlen 설정값 이내에 포함됩니다. '.format(np.sum(num_tokens < max_tokens) / len(num_tokens)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "55af90e7",
   "metadata": {},
   "source": [
    "### 패딩 추가"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 685,
   "id": "585de577",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(146182, 41)\n",
      "(49157, 41)\n"
     ]
    }
   ],
   "source": [
    "X_train = tf.keras.preprocessing.sequence.pad_sequences(X_train,\n",
    "                                                        value=word_to_index[\"<PAD>\"],\n",
    "                                                        padding='pre',\n",
    "                                                        maxlen=maxlen)\n",
    "\n",
    "X_test = tf.keras.preprocessing.sequence.pad_sequences(X_test,\n",
    "                                                       value=word_to_index[\"<PAD>\"],\n",
    "                                                       padding='pre',\n",
    "                                                       maxlen=maxlen)\n",
    "\n",
    "print(X_train.shape)\n",
    "print(X_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "789851e5",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "# 4. 모델 구성 및 validation set 구성"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 686,
   "id": "e3f32ffd",
   "metadata": {},
   "outputs": [],
   "source": [
    "VOCAB_SIZE = 10000\n",
    "WORD_VECTOR_DIM = 10"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1ef79c3c",
   "metadata": {},
   "source": [
    "### RNN 모델 구성"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 687,
   "id": "caf16517",
   "metadata": {},
   "outputs": [],
   "source": [
    "def getRNN_model(vocab_size, word_vector_dim, embedding_layer=False):\n",
    "    if not embedding_layer:\n",
    "        embedding_layer = tf.keras.layers.Embedding(vocab_size, word_vector_dim, input_shape=(None,))\n",
    "        \n",
    "    model = tf.keras.Sequential()\n",
    "    model.add(embedding_layer)\n",
    "    model.add(tf.keras.layers.LSTM(8))\n",
    "    model.add(tf.keras.layers.Dense(8, activation='relu'))\n",
    "    model.add(tf.keras.layers.Dense(1, activation='sigmoid'))\n",
    "\n",
    "    model.summary()\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0f7cb225",
   "metadata": {},
   "source": [
    "### 1-D CNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 688,
   "id": "acc54259",
   "metadata": {},
   "outputs": [],
   "source": [
    "def getCNN_model(vocab_size, word_vector_dim, embedding_layer=False):\n",
    "    if not embedding_layer:\n",
    "        embedding_layer = tf.keras.layers.Embedding(vocab_size, word_vector_dim, input_shape=(None,))\n",
    "        \n",
    "    model = tf.keras.Sequential()\n",
    "    model.add(embedding_layer)\n",
    "    model.add(tf.keras.layers.Conv1D(16, 7, activation='relu'))\n",
    "    model.add(tf.keras.layers.MaxPooling1D(5))\n",
    "    model.add(tf.keras.layers.Conv1D(16, 7, activation='relu'))\n",
    "    model.add(tf.keras.layers.GlobalMaxPooling1D())\n",
    "    model.add(tf.keras.layers.Dense(8, activation='relu'))\n",
    "    model.add(tf.keras.layers.Dense(1, activation='sigmoid'))\n",
    "\n",
    "    model.summary()\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f62d6e44",
   "metadata": {},
   "source": [
    "### GlobalMaxPooling1D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 689,
   "id": "c611eaa1",
   "metadata": {},
   "outputs": [],
   "source": [
    "def getGlobalMaxPooling_model(vocab_size, word_vector_dim, embedding_layer=False):\n",
    "    if not embedding_layer:\n",
    "        embedding_layer = tf.keras.layers.Embedding(vocab_size, word_vector_dim, input_shape=(None,))\n",
    "\n",
    "    model = tf.keras.Sequential()\n",
    "    model.add(embedding_layer)\n",
    "    model.add(tf.keras.layers.GlobalMaxPooling1D())\n",
    "    model.add(tf.keras.layers.Dense(8, activation='relu'))\n",
    "    model.add(tf.keras.layers.Dense(1, activation='sigmoid'))\n",
    "\n",
    "    model.summary()\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "08b4c6ac",
   "metadata": {},
   "source": [
    "### Validation set 구성"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 690,
   "id": "8be57755",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(97025, 41)\n",
      "(97025,)\n"
     ]
    }
   ],
   "source": [
    "val_size = len(X_test) # 테스트셋 사이즈와 동일하게 validation set 구성\n",
    "\n",
    "X_val = X_train[:val_size]\n",
    "y_val = y_train[:val_size]\n",
    "\n",
    "partial_X_train = X_train[val_size:]\n",
    "partial_y_train = y_train[val_size:]\n",
    "\n",
    "print(partial_X_train.shape)\n",
    "print(partial_y_train.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "189b2c3d",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "# 5. 모델 훈련 개시"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e57305ec",
   "metadata": {},
   "source": [
    "### RNN model 훈련"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 691,
   "id": "9def2abb",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_84\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_83 (Embedding)     (None, None, 10)          100000    \n",
      "_________________________________________________________________\n",
      "lstm_39 (LSTM)               (None, 8)                 608       \n",
      "_________________________________________________________________\n",
      "dense_154 (Dense)            (None, 8)                 72        \n",
      "_________________________________________________________________\n",
      "dense_155 (Dense)            (None, 1)                 9         \n",
      "=================================================================\n",
      "Total params: 100,689\n",
      "Trainable params: 100,689\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/20\n",
      "190/190 [==============================] - 3s 8ms/step - loss: 0.5480 - accuracy: 0.7567 - val_loss: 0.4000 - val_accuracy: 0.8338\n",
      "Epoch 2/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.3657 - accuracy: 0.8474 - val_loss: 0.3615 - val_accuracy: 0.8460\n",
      "Epoch 3/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.3303 - accuracy: 0.8640 - val_loss: 0.3548 - val_accuracy: 0.8476\n",
      "Epoch 4/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.3132 - accuracy: 0.8717 - val_loss: 0.3571 - val_accuracy: 0.8469\n",
      "Epoch 5/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.3026 - accuracy: 0.8774 - val_loss: 0.3590 - val_accuracy: 0.8462\n",
      "Epoch 6/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2952 - accuracy: 0.8808 - val_loss: 0.3639 - val_accuracy: 0.8449\n",
      "Epoch 7/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2880 - accuracy: 0.8850 - val_loss: 0.3739 - val_accuracy: 0.8420\n",
      "Epoch 8/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2808 - accuracy: 0.8881 - val_loss: 0.3698 - val_accuracy: 0.8458\n",
      "Epoch 9/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2736 - accuracy: 0.8919 - val_loss: 0.3761 - val_accuracy: 0.8438\n",
      "Epoch 10/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2675 - accuracy: 0.8950 - val_loss: 0.3767 - val_accuracy: 0.8452\n",
      "Epoch 11/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2588 - accuracy: 0.8993 - val_loss: 0.3859 - val_accuracy: 0.8427\n",
      "Epoch 12/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2528 - accuracy: 0.9025 - val_loss: 0.3855 - val_accuracy: 0.8432\n",
      "Epoch 13/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2453 - accuracy: 0.9065 - val_loss: 0.3933 - val_accuracy: 0.8426\n",
      "Epoch 14/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2384 - accuracy: 0.9097 - val_loss: 0.4033 - val_accuracy: 0.8422\n",
      "Epoch 15/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2324 - accuracy: 0.9122 - val_loss: 0.4042 - val_accuracy: 0.8398\n",
      "Epoch 16/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2259 - accuracy: 0.9153 - val_loss: 0.4172 - val_accuracy: 0.8389\n",
      "Epoch 17/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2194 - accuracy: 0.9183 - val_loss: 0.4205 - val_accuracy: 0.8379\n",
      "Epoch 18/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2141 - accuracy: 0.9215 - val_loss: 0.4282 - val_accuracy: 0.8387\n",
      "Epoch 19/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2077 - accuracy: 0.9245 - val_loss: 0.4362 - val_accuracy: 0.8336\n",
      "Epoch 20/20\n",
      "190/190 [==============================] - 1s 6ms/step - loss: 0.2041 - accuracy: 0.9254 - val_loss: 0.4416 - val_accuracy: 0.8356\n"
     ]
    }
   ],
   "source": [
    "RNN_model = getRNN_model(VOCAB_SIZE, WORD_VECTOR_DIM)\n",
    "\n",
    "RNN_model.compile(optimizer='adam',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "epochs=20\n",
    "\n",
    "RNN_history = RNN_model.fit(partial_X_train,\n",
    "                    partial_y_train,\n",
    "                    epochs=epochs,\n",
    "                    batch_size=512,\n",
    "                    validation_data=(X_val, y_val),\n",
    "                    verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 692,
   "id": "39127f98",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1537/1537 - 3s - loss: 0.4471 - accuracy: 0.8327\n",
      "[0.44709113240242004, 0.8326992988586426]\n"
     ]
    }
   ],
   "source": [
    "RNN_results = RNN_model.evaluate(X_test,  y_test, verbose=2)\n",
    "\n",
    "print(RNN_results)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "41c33496",
   "metadata": {},
   "source": [
    "### CNN model 훈련"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 693,
   "id": "9fd3702e",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_85\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_84 (Embedding)     (None, None, 10)          100000    \n",
      "_________________________________________________________________\n",
      "conv1d_76 (Conv1D)           (None, None, 16)          1136      \n",
      "_________________________________________________________________\n",
      "max_pooling1d_38 (MaxPooling (None, None, 16)          0         \n",
      "_________________________________________________________________\n",
      "conv1d_77 (Conv1D)           (None, None, 16)          1808      \n",
      "_________________________________________________________________\n",
      "global_max_pooling1d_38 (Glo (None, 16)                0         \n",
      "_________________________________________________________________\n",
      "dense_156 (Dense)            (None, 8)                 136       \n",
      "_________________________________________________________________\n",
      "dense_157 (Dense)            (None, 1)                 9         \n",
      "=================================================================\n",
      "Total params: 103,089\n",
      "Trainable params: 103,089\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/20\n",
      "190/190 [==============================] - 2s 6ms/step - loss: 0.5210 - accuracy: 0.7367 - val_loss: 0.3719 - val_accuracy: 0.8371\n",
      "Epoch 2/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.3426 - accuracy: 0.8538 - val_loss: 0.3494 - val_accuracy: 0.8470\n",
      "Epoch 3/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.3085 - accuracy: 0.8717 - val_loss: 0.3477 - val_accuracy: 0.8484\n",
      "Epoch 4/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.2848 - accuracy: 0.8839 - val_loss: 0.3498 - val_accuracy: 0.8489\n",
      "Epoch 5/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.2617 - accuracy: 0.8952 - val_loss: 0.3551 - val_accuracy: 0.8492\n",
      "Epoch 6/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.2350 - accuracy: 0.9083 - val_loss: 0.3700 - val_accuracy: 0.8476\n",
      "Epoch 7/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.2082 - accuracy: 0.9210 - val_loss: 0.3907 - val_accuracy: 0.8442\n",
      "Epoch 8/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.1802 - accuracy: 0.9343 - val_loss: 0.4251 - val_accuracy: 0.8407\n",
      "Epoch 9/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.1555 - accuracy: 0.9454 - val_loss: 0.4549 - val_accuracy: 0.8379\n",
      "Epoch 10/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.1328 - accuracy: 0.9552 - val_loss: 0.4924 - val_accuracy: 0.8326\n",
      "Epoch 11/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.1148 - accuracy: 0.9624 - val_loss: 0.5233 - val_accuracy: 0.8296\n",
      "Epoch 12/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.1003 - accuracy: 0.9685 - val_loss: 0.5624 - val_accuracy: 0.8272\n",
      "Epoch 13/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.0886 - accuracy: 0.9730 - val_loss: 0.5919 - val_accuracy: 0.8262\n",
      "Epoch 14/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.0796 - accuracy: 0.9758 - val_loss: 0.6428 - val_accuracy: 0.8228\n",
      "Epoch 15/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.0711 - accuracy: 0.9795 - val_loss: 0.6738 - val_accuracy: 0.8221\n",
      "Epoch 16/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.0659 - accuracy: 0.9809 - val_loss: 0.7111 - val_accuracy: 0.8201\n",
      "Epoch 17/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.0603 - accuracy: 0.9822 - val_loss: 0.7533 - val_accuracy: 0.8192\n",
      "Epoch 18/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.0558 - accuracy: 0.9836 - val_loss: 0.7861 - val_accuracy: 0.8178\n",
      "Epoch 19/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.0514 - accuracy: 0.9853 - val_loss: 0.8158 - val_accuracy: 0.8158\n",
      "Epoch 20/20\n",
      "190/190 [==============================] - 1s 5ms/step - loss: 0.0478 - accuracy: 0.9863 - val_loss: 0.8427 - val_accuracy: 0.8158\n"
     ]
    }
   ],
   "source": [
    "# 1-D CNN model 학습\n",
    "CNN_model = getCNN_model(VOCAB_SIZE, WORD_VECTOR_DIM)\n",
    "\n",
    "CNN_model.compile(optimizer='adam',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "epochs=20\n",
    "\n",
    "CNN_history = CNN_model.fit(partial_X_train,\n",
    "                    partial_y_train,\n",
    "                    epochs=epochs,\n",
    "                    batch_size=512,\n",
    "                    validation_data=(X_val, y_val),\n",
    "                    verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 694,
   "id": "b30870f9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1537/1537 - 2s - loss: 0.8741 - accuracy: 0.8103\n",
      "[0.8741212487220764, 0.8103220462799072]\n"
     ]
    }
   ],
   "source": [
    "CNN_results = CNN_model.evaluate(X_test,  y_test, verbose=2)\n",
    "\n",
    "print(CNN_results)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dee4159c",
   "metadata": {},
   "source": [
    "### GlobalMaxPooling1D model 훈련"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 695,
   "id": "b73a04bb",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_86\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_85 (Embedding)     (None, None, 10)          100000    \n",
      "_________________________________________________________________\n",
      "global_max_pooling1d_39 (Glo (None, 10)                0         \n",
      "_________________________________________________________________\n",
      "dense_158 (Dense)            (None, 8)                 88        \n",
      "_________________________________________________________________\n",
      "dense_159 (Dense)            (None, 1)                 9         \n",
      "=================================================================\n",
      "Total params: 100,097\n",
      "Trainable params: 100,097\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.6415 - accuracy: 0.7014 - val_loss: 0.5345 - val_accuracy: 0.7968\n",
      "Epoch 2/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.4408 - accuracy: 0.8197 - val_loss: 0.3918 - val_accuracy: 0.8269\n",
      "Epoch 3/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.3613 - accuracy: 0.8456 - val_loss: 0.3695 - val_accuracy: 0.8356\n",
      "Epoch 4/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.3303 - accuracy: 0.8600 - val_loss: 0.3636 - val_accuracy: 0.8391\n",
      "Epoch 5/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.3090 - accuracy: 0.8701 - val_loss: 0.3631 - val_accuracy: 0.8406\n",
      "Epoch 6/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2926 - accuracy: 0.8779 - val_loss: 0.3665 - val_accuracy: 0.8400\n",
      "Epoch 7/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2794 - accuracy: 0.8842 - val_loss: 0.3711 - val_accuracy: 0.8391\n",
      "Epoch 8/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2678 - accuracy: 0.8901 - val_loss: 0.3773 - val_accuracy: 0.8379\n",
      "Epoch 9/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2580 - accuracy: 0.8948 - val_loss: 0.3840 - val_accuracy: 0.8369\n",
      "Epoch 10/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2493 - accuracy: 0.8995 - val_loss: 0.3924 - val_accuracy: 0.8355\n",
      "Epoch 11/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2416 - accuracy: 0.9028 - val_loss: 0.4006 - val_accuracy: 0.8341\n",
      "Epoch 12/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2349 - accuracy: 0.9064 - val_loss: 0.4077 - val_accuracy: 0.8327\n",
      "Epoch 13/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2287 - accuracy: 0.9095 - val_loss: 0.4153 - val_accuracy: 0.8316\n",
      "Epoch 14/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2232 - accuracy: 0.9115 - val_loss: 0.4234 - val_accuracy: 0.8308\n",
      "Epoch 15/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2182 - accuracy: 0.9145 - val_loss: 0.4301 - val_accuracy: 0.8313\n",
      "Epoch 16/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2136 - accuracy: 0.9169 - val_loss: 0.4372 - val_accuracy: 0.8298\n",
      "Epoch 17/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2093 - accuracy: 0.9193 - val_loss: 0.4462 - val_accuracy: 0.8286\n",
      "Epoch 18/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2056 - accuracy: 0.9206 - val_loss: 0.4522 - val_accuracy: 0.8282\n",
      "Epoch 19/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.2018 - accuracy: 0.9225 - val_loss: 0.4586 - val_accuracy: 0.8269\n",
      "Epoch 20/20\n",
      "190/190 [==============================] - 1s 4ms/step - loss: 0.1983 - accuracy: 0.9242 - val_loss: 0.4652 - val_accuracy: 0.8269\n"
     ]
    }
   ],
   "source": [
    "# GlobalMaxPooling1D model 학습\n",
    "globalMaxPooling_model = getGlobalMaxPooling_model(VOCAB_SIZE, WORD_VECTOR_DIM)\n",
    "\n",
    "globalMaxPooling_model.compile(optimizer='adam',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "epochs=20\n",
    "\n",
    "globalMaxPooling_history = globalMaxPooling_model.fit(partial_X_train,\n",
    "                    partial_y_train,\n",
    "                    epochs=epochs,\n",
    "                    batch_size=512,\n",
    "                    validation_data=(X_val, y_val),\n",
    "                    verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 696,
   "id": "144c21d0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1537/1537 - 2s - loss: 0.4735 - accuracy: 0.8231\n",
      "[0.4734635651111603, 0.8230567574501038]\n"
     ]
    }
   ],
   "source": [
    "globalMaxPooling_results = globalMaxPooling_model.evaluate(X_test,  y_test, verbose=2)\n",
    "\n",
    "print(globalMaxPooling_results)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ca497b12",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "# 6. Loss, Accuracy 그래프 시각화"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1fb66f54",
   "metadata": {},
   "source": [
    "### RNN model - Loss, Accuracy 시각화"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 697,
   "id": "8bd3363b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x1080 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def visualize(name, history, plt, height, idx=1):\n",
    "    history_dict = history.history\n",
    "\n",
    "    acc = history_dict['accuracy']\n",
    "    val_acc = history_dict['val_accuracy']\n",
    "    loss = history_dict['loss']\n",
    "    val_loss = history_dict['val_loss']\n",
    "\n",
    "    epochs = range(1, len(acc) + 1)\n",
    "\n",
    "\n",
    "    ax1 = fig.add_subplot(height,2,idx)\n",
    "    plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "    plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "\n",
    "    plt.title(name + ' Training and validation loss')\n",
    "    plt.xlabel('Epochs')\n",
    "    plt.ylabel('Loss')\n",
    "    plt.legend()\n",
    "\n",
    "    ax2 = fig.add_subplot(height,2,idx+1)\n",
    "    plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "    plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "    plt.title(name + ' Training and validation accuracy')\n",
    "    plt.xlabel('Epochs')\n",
    "    plt.ylabel('Accuracy')\n",
    "    plt.legend()\n",
    "\n",
    "    \n",
    "histories = {'RNN': RNN_history, 'CNN': CNN_history, 'GlobalMaxPooling': globalMaxPooling_history}\n",
    "\n",
    "fig = plt.figure(figsize=(12,15))\n",
    "idx = 0\n",
    "for name, h in histories.items():\n",
    "    visualize(name, h, plt, 3, (idx*2)+1)\n",
    "    idx += 1\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c6d533bc",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "# 7. 학습된 Embedding 레이어 분석"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 698,
   "id": "9848a9d0",
   "metadata": {},
   "outputs": [],
   "source": [
    "embedding_layer = globalMaxPooling_model.layers[0]\n",
    "weights = embedding_layer.get_weights()[0]\n",
    "# print(weights.shape)\n",
    "\n",
    "word2vec_file_path = os.getenv('HOME')+'/aiffel/sentiment_classification/data/word2vec.txt'\n",
    "f = open(word2vec_file_path, 'w')\n",
    "f.write('{} {}\\n'.format(VOCAB_SIZE-4, WORD_VECTOR_DIM))\n",
    "\n",
    "vectors = globalMaxPooling_model.get_weights()[0]\n",
    "for i in range(4,VOCAB_SIZE):\n",
    "    f.write('{} {}\\n'.format(index_to_word[i], ' '.join(map(str, list(vectors[i, :])))))\n",
    "f.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 699,
   "id": "c2071cd0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 0.06348492, -0.18699528, -0.2444353 , -0.01120742, -0.04669213,\n",
       "       -0.13194396, -0.18469626, -0.17511238, -0.19390588, -0.01857155],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 699,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from gensim.models.keyedvectors import Word2VecKeyedVectors\n",
    "\n",
    "word_vectors = Word2VecKeyedVectors.load_word2vec_format(word2vec_file_path, binary=False)\n",
    "vector = word_vectors['사람']\n",
    "vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 700,
   "id": "bab6e2be",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('됨', 0.9737274050712585),\n",
       " ('취미', 0.9631553888320923),\n",
       " ('여', 0.9628081917762756),\n",
       " ('위안', 0.954928994178772),\n",
       " ('모르', 0.9491946697235107),\n",
       " ('유명', 0.9489961862564087),\n",
       " ('쪽', 0.9329440593719482),\n",
       " ('관심', 0.9303947687149048),\n",
       " ('발레', 0.9292678833007812),\n",
       " ('가정', 0.9282903671264648)]"
      ]
     },
     "execution_count": 700,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "word_vectors.similar_by_word(\"사람\") # \"사랑\" 단어와 가장 유사한 단어들을 출력"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ed9cde23",
   "metadata": {},
   "source": [
    "그나마 가장 나은 성능을 가진 모델의 워드벡터로 유사도를 분석해봤으나, 좋은 결과를 내지 않았습니다."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5fc31a78",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "# 8. 한국어 Word2Vec 임베딩 활용하여 성능 개선"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 701,
   "id": "95acb4c8",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 0.7412252 , -1.1187236 ,  2.0681474 ,  1.05749   , -1.3211275 ,\n",
       "       -2.414994  , -1.133681  ,  1.5452526 , -0.4543698 ,  0.02380616,\n",
       "        0.12370981, -0.8839078 , -2.9051683 , -0.34940982,  1.3418512 ,\n",
       "        2.8225965 ,  1.0794274 ,  0.4287113 , -1.9613483 ,  0.4262147 ,\n",
       "       -0.40650806, -0.34994036, -0.8416261 ,  0.52574843, -0.29543158,\n",
       "       -0.5593075 ,  0.84687376,  0.0889315 , -2.5240638 ,  0.46828857,\n",
       "        2.8982353 , -1.6508341 , -0.37446976, -2.2825975 ,  0.7083552 ,\n",
       "       -0.59711653,  0.2839199 , -1.0247895 ,  2.3036754 ,  0.45918384,\n",
       "       -0.83396995,  0.85293204,  0.972802  , -0.5457918 , -0.93798906,\n",
       "       -0.8009529 ,  2.160266  ,  0.10052549, -0.16650781, -0.07592812,\n",
       "       -0.45869842,  1.3352628 ,  2.3284414 ,  0.7132327 , -2.4690466 ,\n",
       "        0.05213251,  0.38745207, -0.2773064 , -0.67375   , -0.59059584,\n",
       "        1.0310122 ,  1.9923794 , -0.94680166, -1.3167765 , -1.4717319 ,\n",
       "        0.97447944,  0.8103551 ,  0.8915873 ,  0.7474684 , -0.00637981,\n",
       "        0.7444418 ,  0.7939502 , -1.0944605 ,  1.392656  ,  0.37654135,\n",
       "       -1.2034026 ,  1.0072207 , -0.40544778,  0.44212767, -3.0634484 ,\n",
       "        1.1875361 , -0.12091239,  1.4954131 ,  0.36426282, -1.0021433 ,\n",
       "        0.11045299, -0.8199569 ,  0.73303777,  0.26993483,  0.5307779 ,\n",
       "        3.26409   , -1.0755528 , -1.1228558 ,  1.6779242 ,  1.790663  ,\n",
       "       -0.2767947 ,  1.4767654 ,  0.08618896,  0.5165028 ,  2.7088833 ,\n",
       "       -0.46314278,  0.99281746,  0.39835286,  0.88094866, -0.6806667 ,\n",
       "        1.6146507 ,  0.40473613,  0.4831634 ,  0.8650192 , -2.2554002 ,\n",
       "        1.4073339 ,  0.29678002,  0.312536  , -0.24908657, -0.63724786,\n",
       "        0.01833764,  0.28282273,  0.5918058 , -0.4271871 , -0.36062592,\n",
       "        0.28868324, -1.5452292 , -2.195867  , -3.35319   ,  0.90652514,\n",
       "        0.69410455, -2.4260206 ,  0.62638485,  1.4539987 ,  0.66689837,\n",
       "       -0.43008897, -1.0371583 ,  1.7439222 ,  2.1849706 ,  2.809909  ,\n",
       "       -0.3376442 ,  0.6059654 , -0.11268281,  0.5642306 , -2.0597198 ,\n",
       "        0.2558234 , -0.8461995 ,  1.9621457 ,  2.4020708 ,  0.10118579,\n",
       "        0.7569208 , -0.52231586,  0.69310755, -0.12435903,  1.7258118 ,\n",
       "       -0.47222325, -0.11036941,  0.09247034, -0.24319325,  0.26812062,\n",
       "       -1.8018712 ,  0.31583163,  0.14499973, -0.6330405 ,  1.0520003 ,\n",
       "       -1.4853543 ,  0.34385768,  0.5239166 , -2.2002435 , -2.701687  ,\n",
       "       -1.4307605 ,  0.77635235,  2.863714  , -0.73031855,  1.1225786 ,\n",
       "       -1.8527915 , -2.1887116 , -1.2744621 , -0.53644854,  1.459293  ,\n",
       "       -0.47653604,  0.5469871 , -0.22993626,  1.1055657 ,  0.47264385,\n",
       "       -1.3790879 ,  2.063961  ,  2.1002257 , -0.16041307, -1.7497696 ,\n",
       "        1.8129411 , -0.30766737, -0.9470096 , -1.6739731 ,  0.02093687,\n",
       "       -0.26385018,  0.18818851, -0.520275  ,  0.08425661,  1.003671  ,\n",
       "        0.97272223,  0.85132086,  1.3672657 , -0.7964044 , -2.9975104 ],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 701,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import gensim\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\", category=DeprecationWarning) \n",
    "\n",
    "word2vec_path = '/aiffel/aiffel/sentiment_classification/data/ko.bin'\n",
    "word2vec = gensim.models.Word2Vec.load(word2vec_path)\n",
    "vector = word2vec['사람']\n",
    "vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 702,
   "id": "2a90c348",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2203153"
      ]
     },
     "execution_count": 702,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# dir(word2vec)\n",
    "word2vec.corpus_count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 703,
   "id": "fe95c811",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('젊은이', 0.6494427919387817),\n",
       " ('여인', 0.6287257671356201),\n",
       " ('백성', 0.6063710451126099),\n",
       " ('포졸', 0.6043275594711304),\n",
       " ('죄인', 0.5960500836372375),\n",
       " ('선비', 0.5868039131164551),\n",
       " ('부녀자', 0.5654411315917969),\n",
       " ('죄수', 0.5639811754226685),\n",
       " ('구경꾼', 0.5620019435882568),\n",
       " ('손님', 0.5589558482170105)]"
      ]
     },
     "execution_count": 703,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 메모리를 다소 많이 소비하는 작업이니 유의해 주세요.\n",
    "word2vec.similar_by_word(\"사람\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d7275035",
   "metadata": {},
   "source": [
    "### 한국어 Word2Vec으로 임베딩 레이어 교체하기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 704,
   "id": "7e93fe1d",
   "metadata": {},
   "outputs": [],
   "source": [
    "vocab_size = 10000\n",
    "word_vector_dim = 200\n",
    "embedding_matrix = np.random.rand(vocab_size, word_vector_dim)\n",
    "\n",
    "for i in range(4, vocab_size):\n",
    "    if index_to_word[i] in word2vec:\n",
    "        embedding_matrix[i] = word2vec[index_to_word[i]]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e9c56579",
   "metadata": {},
   "source": [
    "### 모델 다시 작성 & 학습 & 평가"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 726,
   "id": "56503188",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_93\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_92 (Embedding)     (None, 41, 200)           2000000   \n",
      "_________________________________________________________________\n",
      "lstm_44 (LSTM)               (None, 8)                 6688      \n",
      "_________________________________________________________________\n",
      "dense_172 (Dense)            (None, 8)                 72        \n",
      "_________________________________________________________________\n",
      "dense_173 (Dense)            (None, 1)                 9         \n",
      "=================================================================\n",
      "Total params: 2,006,769\n",
      "Trainable params: 2,006,769\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.initializers import Constant\n",
    "\n",
    "vocab_size = 10000\n",
    "word_vector_dim = 200\n",
    "embedding_layer = tf.keras.layers.Embedding(vocab_size, \n",
    "                                 word_vector_dim,\n",
    "                                 embeddings_initializer=Constant(embedding_matrix),\n",
    "                                 input_length=maxlen, \n",
    "                                 trainable=True)\n",
    "\n",
    "\n",
    "model = getRNN_model(vocab_size, word_vector_dim, embedding_layer)\n",
    "# model = getCNN_model(vocab_size, word_vector_dim, embedding_layer)\n",
    "# model = getGlobalMaxPooling_model(vocab_size, word_vector_dim, embedding_layer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 727,
   "id": "01ee2c08",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/4\n",
      "190/190 [==============================] - 3s 9ms/step - loss: 0.5370 - accuracy: 0.7265 - val_loss: 0.4233 - val_accuracy: 0.8106\n",
      "Epoch 2/4\n",
      "190/190 [==============================] - 1s 7ms/step - loss: 0.3784 - accuracy: 0.8364 - val_loss: 0.3688 - val_accuracy: 0.8397\n",
      "Epoch 3/4\n",
      "190/190 [==============================] - 1s 7ms/step - loss: 0.3298 - accuracy: 0.8633 - val_loss: 0.3587 - val_accuracy: 0.8468\n",
      "Epoch 4/4\n",
      "190/190 [==============================] - 1s 7ms/step - loss: 0.3025 - accuracy: 0.8773 - val_loss: 0.3509 - val_accuracy: 0.8505\n"
     ]
    }
   ],
   "source": [
    "# 학습의 진행\n",
    "model.compile(optimizer='adam',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "              \n",
    "epochs=4\n",
    "\n",
    "history = model.fit(partial_X_train,\n",
    "                    partial_y_train,\n",
    "                    epochs=epochs,\n",
    "                    batch_size=512,\n",
    "                    validation_data=(X_val, y_val),\n",
    "                    verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 728,
   "id": "66244597",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1537/1537 - 3s - loss: 0.3623 - accuracy: 0.8462\n",
      "[0.3623000681400299, 0.8462477326393127]\n"
     ]
    }
   ],
   "source": [
    "# 테스트셋을 통한 모델 평가\n",
    "results = model.evaluate(X_test, y_test, verbose=2)\n",
    "\n",
    "print(results)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ccdf7e8d",
   "metadata": {},
   "source": [
    "# 시각화"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 729,
   "id": "4f3f0ee2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.clf()   # 그림을 초기화합니다\n",
    "fig = plt.figure(figsize=(12,5))\n",
    "\n",
    "visualize('New Word2Vec', history, plt, 1)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6275c78f",
   "metadata": {},
   "source": [
    "# Accuracy, loss 시도한 기록들\n",
    "\n",
    "1. global\n",
    "```\n",
    "    # 1. 첫 시도\n",
    "    1537/1537 - 2s - loss: 0.4470 - accuracy: 0.8163\n",
    "    [0.4470440745353699, 0.8163435459136963]\n",
    "    \n",
    "    # 2. 문장길이 키우기\n",
    "    1537/1537 - 2s - loss: 0.4235 - accuracy: 0.8212\n",
    "    [0.4234999418258667, 0.8212462067604065]\n",
    "    \n",
    "    # 3. 한글만 filter, 문장 길이 키우기\n",
    "    1537/1537 - 2s - loss: 0.4450 - accuracy: 0.8122\n",
    "    [0.4450375437736511, 0.8121732473373413]\n",
    "```\n",
    "\n",
    "2. RNN\n",
    "```\n",
    "    # 4. 다른 모델로 적용해보기\n",
    "    1537/1537 - 3s - loss: 0.8146 - accuracy: 0.8216\n",
    "    [0.8145668506622314, 0.8216123580932617]\n",
    "    \n",
    "    # 특수문자 제거 후 epoch(6)\n",
    "    1537/1537 - 3s - loss: 0.3744 - accuracy: 0.8407\n",
    "    \n",
    "    # epoch(10)\n",
    "    1537/1537 - 3s - loss: 0.3823 - accuracy: 0.8431\n",
    "    \n",
    "    # 모두 제거 + epoch(5)\n",
    "    1537/1537 - 3s - loss: 0.3824 - accuracy: 0.8347\n",
    "    \n",
    "    # 아무것도 안하고 + epoch(6)\n",
    "    1537/1537 - 3s - loss: 0.3723 - accuracy: 0.8444\n",
    "    \n",
    "    # filter 다 적용 + epoch(7)\n",
    "    1537/1537 - 3s - loss: 0.3803 - accuracy: 0.8431\n",
    "    \n",
    "    # filter 안하고 + epoch(6) + 사전 6000개\n",
    "    1537/1537 - 3s - loss: 0.3642 - accuracy: 0.8428\n",
    "```\n",
    "\n",
    "3. CNN\n",
    "```\n",
    "    1537/1537 - 3s - loss: 0.3938 - accuracy: 0.8374\n",
    "    \n",
    "    # 특수문자 제거 후\n",
    "    1537/1537 - 2s - loss: 0.4231 - accuracy: 0.8164\n",
    "    \n",
    "    # 아무것도 안하고 + epoch(5)\n",
    "    1537/1537 - 2s - loss: 0.3764 - accuracy: 0.8401\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "830e661c",
   "metadata": {},
   "source": [
    "# 회고\n",
    "\n",
    "### 1. 어려웠던 점\n",
    "- 데이터 전처리를 해도, 사전 크기를 늘려봐도, 문장 최대 길이를 늘려봐도... 정확도,손실에 큰 개선을 이뤄내지 못해 어려웠습니다.\n",
    "\n",
    "\n",
    "### 2. 시도한 것들\n",
    "- word_vector_dimension 사이즈를 어떤걸로 설정해야할까?\n",
    "    - 처음에는 적정 사이즈를 어떻게 두면 좋을지 검색해보고, number_of_categories**0.25 이렇게 초기 설정해서 진행했습니다.\n",
    "        - https://stackoverflow.com/questions/48479915/what-is-the-preferred-ratio-between-the-vocabulary-size-and-embedding-dimension\n",
    "    - 이후에는 accuracy가 잘 나오는 값을 찾아서 값을 늘려 학습해보았습니다.\n",
    "- validation dataset 개수는 얼마가 적절할까?\n",
    "    - test dataset size는 0.2가 적절하다는걸 아는데, validation dataset size는 어떻게 구성하는게 좋을지 찾아봤습니다.\n",
    "        - https://stackoverflow.com/questions/13610074/is-there-a-rule-of-thumb-for-how-to-divide-a-dataset-into-training-and-validatio\n",
    "    - validation dataset도 0.2가 적절하다고 합니다. (작은 데이터셋인 경우에는 7:3 또는 6:4도 권장)\n",
    "    - 그러나 현재 train, test dataset은 이미 나눠진 상태였기 때문에, validation size를 test size와 동일하게 구성하였습니다.\n",
    "    \n",
    "- 한글 임베딩 파일을 로드할 때 에러 발생발생했고, 아래 링크에서 문제해결을 찾아 에러를 고쳤습니다.\n",
    "    - https://stackoverflow.com/questions/27659985/error-utf8-codec-cant-decode-byte-0x80-in-position-0-invalid-start-byte\n",
    "- 데이터 전처리 추가\n",
    "    - 영어, 숫자, 특수문자는 한글 텍스트 분석에 적절하지 않다 판단하였고, 데이터 전처리를 통해 데이터에 한글만 유효하도록 수정했습니다. (자음, 모음만 있는 것도 삭제)\n",
    "- 데이터 전처리를 추가하고 난 뒤, \"CNN Negative dimension size caused by subtracting ..\"와 같은 에러가 발생했습니다.\n",
    "    - 문장 최대 길이를 2 -> 2.5 가중치를 늘려 오류를 해결했습니다.\n",
    "        - max_tokens = np.mean(num_tokens) + 2.5 * np.std(num_tokens)\n",
    "- accuracy, loss가 잘 나오는 범위를 찾아 epoch를 설정했습니다.\n",
    "\n",
    "\n",
    "### 3. 아쉬운 점\n",
    "```\n",
    "가장 성능이 괜찮아 보였던 GlobalMaxPooling1D 기준으로 해보기도 했고\n",
    "데이터 전처리를 하거나, 사전 크기를 늘려봐도, 문장 최대 길이를 늘려봐도... 정확도,손실에 큰 개선을 이뤄내지 못했습니다. \n",
    "그래도 데이터 전처리를 더 정교하게 했다면 어땠을지... 이 부분이 아쉽습니다.\n",
    "```\n"
   ]
  }
 ],
 "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
}