8. predict_spam_or_ham_with_LSTM.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 스팸 메일인지 스펨인지 RNN(LSTM)으로 검사하는 예제"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import os\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from keras.models import Model, Sequential\n",
    "from keras.layers import Dense, LSTM, Dropout, Embedding, Input\n",
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.preprocessing import sequence\n",
    "from keras.utils import to_categorical\n",
    "from keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "import keras.backend.tensorflow_backend as K\n",
    "import tensorflow as tf\n",
    "config = tf.ConfigProto()\n",
    "config.gpu_options.allow_growth = True\n",
    "session = tf.Session(config=config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "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>v1</th>\n",
       "      <th>v2</th>\n",
       "      <th>Unnamed: 2</th>\n",
       "      <th>Unnamed: 3</th>\n",
       "      <th>Unnamed: 4</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ham</td>\n",
       "      <td>Go until jurong point, crazy.. Available only ...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ham</td>\n",
       "      <td>Ok lar... Joking wif u oni...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>spam</td>\n",
       "      <td>Free entry in 2 a wkly comp to win FA Cup fina...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ham</td>\n",
       "      <td>U dun say so early hor... U c already then say...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ham</td>\n",
       "      <td>Nah I don't think he goes to usf, he lives aro...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     v1                                                 v2 Unnamed: 2  \\\n",
       "0   ham  Go until jurong point, crazy.. Available only ...        NaN   \n",
       "1   ham                      Ok lar... Joking wif u oni...        NaN   \n",
       "2  spam  Free entry in 2 a wkly comp to win FA Cup fina...        NaN   \n",
       "3   ham  U dun say so early hor... U c already then say...        NaN   \n",
       "4   ham  Nah I don't think he goes to usf, he lives aro...        NaN   \n",
       "\n",
       "  Unnamed: 3 Unnamed: 4  \n",
       "0        NaN        NaN  \n",
       "1        NaN        NaN  \n",
       "2        NaN        NaN  \n",
       "3        NaN        NaN  \n",
       "4        NaN        NaN  "
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv('./dataset/spam.csv', delimiter=',', encoding='latin-1')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "이렇게 필요한 라이브러리를 import 하고 데이터를 가져온다.\n",
    "\n",
    "그리고 head를 쳐보니 이상한 컬럼이 보인다. 저 컬럼은 제거!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 5572 entries, 0 to 5571\n",
      "Data columns (total 2 columns):\n",
      "v1    5572 non-null object\n",
      "v2    5572 non-null object\n",
      "dtypes: object(2)\n",
      "memory usage: 87.1+ KB\n"
     ]
    }
   ],
   "source": [
    "df.drop(['Unnamed: 2', 'Unnamed: 3', 'Unnamed: 4'], axis=1, inplace=True)\n",
    "df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "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>v1</th>\n",
       "      <th>v2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ham</td>\n",
       "      <td>Go until jurong point, crazy.. Available only ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ham</td>\n",
       "      <td>Ok lar... Joking wif u oni...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>spam</td>\n",
       "      <td>Free entry in 2 a wkly comp to win FA Cup fina...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ham</td>\n",
       "      <td>U dun say so early hor... U c already then say...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ham</td>\n",
       "      <td>Nah I don't think he goes to usf, he lives aro...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     v1                                                 v2\n",
       "0   ham  Go until jurong point, crazy.. Available only ...\n",
       "1   ham                      Ok lar... Joking wif u oni...\n",
       "2  spam  Free entry in 2 a wkly comp to win FA Cup fina...\n",
       "3   ham  U dun say so early hor... U c already then say...\n",
       "4   ham  Nah I don't think he goes to usf, he lives aro..."
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "이렇게 하면 이제 v1, v2 두 개의 컬럼만 남게 된다.\n",
    "\n",
    "여기서 스팸하고 스팸이 아닌 것 하고 개수를 세어보자."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'number of ham and spam')"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.countplot(df.v1)\n",
    "plt.xlabel(\"Label\")\n",
    "plt.title(\"number of ham and spam\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "불균형한 데이터셋입니다. 따라서 accuracy도 중요하지만 recall과 precision도 생각하면서 진행하셔야합니다. 여기서는 일단 그냥 해보죠."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'numpy.ndarray'>\n"
     ]
    }
   ],
   "source": [
    "X = df.v2.values\n",
    "y = df.v1.values\n",
    "\n",
    "print(type(X))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "라벨값이 문자값이므로 숫자로 바꿔줍니다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0 0 1 ... 0 0 0]\n"
     ]
    }
   ],
   "source": [
    "le = LabelEncoder()\n",
    "y = le.fit_transform(y)\n",
    "print(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5572,)\n",
      "(5572,)\n"
     ]
    }
   ],
   "source": [
    "print(y.shape)\n",
    "print(X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "shape로 확인했듯이 총 글의 개수는 5572개입니다.\n",
    "\n",
    "여기서 가장 많이 사용한 단어 상위 1000개만 사용할것입니다.\n",
    "\n",
    "그리고 글마다 **길이**가 다르기 때문에 길이도 제한하겠습니다. 150글자로 제한하죠.\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "max_words = 1000\n",
    "max_len = 150\n",
    "\n",
    "#토크나이저로 변환\n",
    "tok = Tokenizer(num_words = max_words)\n",
    "tok.fit_on_texts(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "sequences = tok.texts_to_sequences(X_train)  #텍스트 to 시퀀스를 해야 문자 -> 숫자로 변경이 된다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "25\n",
      "10\n",
      "17\n",
      "4\n",
      "4\n",
      "5\n",
      "19\n",
      "21\n",
      "5\n",
      "14\n",
      "10\n"
     ]
    }
   ],
   "source": [
    "cnt = 0\n",
    "for i in sequences:\n",
    "    print(len(i))\n",
    "    cnt += 1\n",
    "    if cnt > 10:\n",
    "        break"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "이렇게 문자 -> 숫자 작업을 마쳤습니다. 문자에 각 숫자 값은 1이면 제일 많이 나온 단어 첫 번째고 10이면 제일 많이 나온 단어 10번째 라는 것입니다.\n",
    "\n",
    "이제 길이를 제한해서 시퀀스를 만들어봅니다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "sequences_matrix = sequence.pad_sequences(sequences, maxlen=max_len)\n",
    "#문자열 길이를 맞춰서 제한했습니다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0 223\n",
      "  37 568  18 301 107  21 481  10 211   4 109 159 173  44  12 116   7   9\n",
      "  85  25  33  77  78 171]\n",
      "[  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0 103  40   8  26\n",
      "   1 160   6  28 635 461]\n",
      "[  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   1 462   2 148 124   2 211  42  10   2 302\n",
      " 177  11   9 119 482 384]\n",
      "[  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0\n",
      "   0   0 405 246  12 116]\n"
     ]
    }
   ],
   "source": [
    "cnt = 0\n",
    "for i in sequences_matrix:  #이걸로 보면 위쪽에 sequences를 for문으로 출력했ㅇ늘 때랑 똑같은 숫자가 출력. 길이가 안되는 애들은 0으로 되고.\n",
    "    print(i)\n",
    "    cnt += 1\n",
    "    if cnt > 3: break"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "150글자로 제한했기 때문에 150글자가 안되는 애들은 0으로 채워집니다.\n",
    "\n",
    "이제 모델을 만들죠."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "with K.tf_ops.device('device:GPU:0'):\n",
    "    model = Sequential()\n",
    "    #max_words를 50차원에. 즉 1000개의 단어를 50차원에다가. 문장의 길이는 max_len\n",
    "    model.add(Embedding(max_words, 50, input_length=max_len))\n",
    "    model.add(LSTM(64))\n",
    "    model.add(Dense(256, activation='relu'))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(1, activation='sigmoid'))\n",
    "    \n",
    "    model.compile(loss='binary_crossentropy',optimizer='adam',metrics=['accuracy'])\n",
    "    model_dir = './model'\n",
    "    if not os.path.exists(model_dir):\n",
    "        os.mkdir(model_dir)\n",
    "    model_path = model_dir + \"/spam_ham_LSTM.model\"\n",
    "    \n",
    "    checkpoint = ModelCheckpoint(filepath=model_path, monitor='val_loss', verbose=1, save_best_only=True)\n",
    "    early_stopping = EarlyStopping(monitor='val_loss', patience=7)\n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_4 (Embedding)      (None, 150, 50)           50000     \n",
      "_________________________________________________________________\n",
      "lstm_4 (LSTM)                (None, 64)                29440     \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 256)               16640     \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 1)                 257       \n",
      "=================================================================\n",
      "Total params: 96,337\n",
      "Trainable params: 96,337\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 3565 samples, validate on 892 samples\n",
      "Epoch 1/10\n",
      "3565/3565 [==============================] - 8s 2ms/step - loss: 0.5002 - acc: 0.8550 - val_loss: 0.3313 - val_acc: 0.8475\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 0.33130, saving model to ./model/spam_ham_LSTM.model\n",
      "Epoch 2/10\n",
      "3565/3565 [==============================] - 7s 2ms/step - loss: 0.1919 - acc: 0.9119 - val_loss: 0.1218 - val_acc: 0.9720\n",
      "\n",
      "Epoch 00002: val_loss improved from 0.33130 to 0.12176, saving model to ./model/spam_ham_LSTM.model\n",
      "Epoch 3/10\n",
      "3565/3565 [==============================] - 7s 2ms/step - loss: 0.0622 - acc: 0.9846 - val_loss: 0.0581 - val_acc: 0.9809\n",
      "\n",
      "Epoch 00003: val_loss improved from 0.12176 to 0.05813, saving model to ./model/spam_ham_LSTM.model\n",
      "Epoch 4/10\n",
      "3565/3565 [==============================] - 7s 2ms/step - loss: 0.0327 - acc: 0.9896 - val_loss: 0.0570 - val_acc: 0.9798\n",
      "\n",
      "Epoch 00004: val_loss improved from 0.05813 to 0.05698, saving model to ./model/spam_ham_LSTM.model\n",
      "Epoch 5/10\n",
      "3565/3565 [==============================] - 7s 2ms/step - loss: 0.0224 - acc: 0.9938 - val_loss: 0.0670 - val_acc: 0.9798\n",
      "\n",
      "Epoch 00005: val_loss did not improve from 0.05698\n",
      "Epoch 6/10\n",
      "3565/3565 [==============================] - 6s 2ms/step - loss: 0.0139 - acc: 0.9969 - val_loss: 0.0650 - val_acc: 0.9776\n",
      "\n",
      "Epoch 00006: val_loss did not improve from 0.05698\n",
      "Epoch 7/10\n",
      "3565/3565 [==============================] - 6s 2ms/step - loss: 0.0102 - acc: 0.9975 - val_loss: 0.0713 - val_acc: 0.9787\n",
      "\n",
      "Epoch 00007: val_loss did not improve from 0.05698\n",
      "Epoch 8/10\n",
      "3565/3565 [==============================] - 6s 2ms/step - loss: 0.0075 - acc: 0.9983 - val_loss: 0.0853 - val_acc: 0.9776\n",
      "\n",
      "Epoch 00008: val_loss did not improve from 0.05698\n",
      "Epoch 9/10\n",
      "3565/3565 [==============================] - 7s 2ms/step - loss: 0.0084 - acc: 0.9978 - val_loss: 0.0901 - val_acc: 0.9809\n",
      "\n",
      "Epoch 00009: val_loss did not improve from 0.05698\n",
      "Epoch 10/10\n",
      "3565/3565 [==============================] - 7s 2ms/step - loss: 0.0043 - acc: 0.9992 - val_loss: 0.0972 - val_acc: 0.9821\n",
      "\n",
      "Epoch 00010: val_loss did not improve from 0.05698\n"
     ]
    }
   ],
   "source": [
    "hist = model.fit(sequences_matrix, y_train, batch_size=128, epochs=10, validation_split=0.2, callbacks=[checkpoint, early_stopping])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1115/1115 [==============================] - 4s 3ms/step\n",
      "정확도 : 0.9830\n"
     ]
    }
   ],
   "source": [
    "sequences_test = tok.texts_to_sequences(X_test)\n",
    "sequences_matrix_test = sequence.pad_sequences(sequences_test, maxlen=max_len)\n",
    "print(\"정확도 : %.4f\" % (model.evaluate(sequences_matrix_test, y_test)[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_vloss = hist.history['val_loss']\n",
    "y_loss = hist.history['loss']\n",
    "\n",
    "x_len = np.arange(len(y_loss))\n",
    "\n",
    "plt.plot(x_len, y_vloss, marker='.', c='red', label='val_set_loss')\n",
    "plt.plot(x_len, y_loss, marker='.', c='blue', label='train_set_oss')\n",
    "plt.legend()\n",
    "plt.xlabel('epochs')\n",
    "plt.ylabel('loss')\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_vloss = hist.history['val_acc']\n",
    "y_loss = hist.history['acc']\n",
    "\n",
    "x_len = np.arange(len(y_loss))\n",
    "\n",
    "plt.plot(x_len, y_vloss, marker='.', c='red', label='val_set_loss')\n",
    "plt.plot(x_len, y_loss, marker='.', c='blue', label='train_set_oss')\n",
    "plt.legend()\n",
    "plt.xlabel('epochs')\n",
    "plt.ylabel('loss')\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "정확도가 이렇게 나오게 됩니다!"
   ]
  },
  {
   "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}