7.2. seoul-120-baseline-rnn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "M4_PHrOVj51Z"
   },
   "source": [
    "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/pytextbook/pytextbook/blob/main/7.2.%20seoul-120-baseline-rnn.ipynb) \n",
    "\n",
    "## RNN (Recurrent Neural Network) 으로 텍스트 분류하기\n",
    "\n",
    "* API Document: https://www.tensorflow.org/api_docs/python/tf/keras/layers/RNN\n",
    "\n",
    "## 라이브러리 로드"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "executionInfo": {
     "elapsed": 439,
     "status": "ok",
     "timestamp": 1672570135870,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "4CcTgpxw6Aai"
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ESivktS1lEm7"
   },
   "source": [
    "## 데이터 미리보기 및 요약"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 583,
     "status": "ok",
     "timestamp": 1672570136446,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "eFKC0rSZqNsO",
    "outputId": "8fa26a3c-be79-4b86-d764-0d9611212fc9"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2645, 5)"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv(\"https://bit.ly/seoul-120-text-csv\")\n",
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 441
    },
    "executionInfo": {
     "elapsed": 6,
     "status": "ok",
     "timestamp": 1672570136447,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "hkHHw0dXZ0JE",
    "outputId": "538d2cc5-b746-47de-a0a8-e8557a05e1c9"
   },
   "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>번호</th>\n",
       "      <th>분류</th>\n",
       "      <th>제목</th>\n",
       "      <th>내용</th>\n",
       "      <th>내용번호</th>\n",
       "      <th>문서</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2645</td>\n",
       "      <td>복지</td>\n",
       "      <td>아빠 육아휴직 장려금</td>\n",
       "      <td>아빠 육아휴직 장려금   업무개요  남성근로자의 육아휴직을 장려하고 양육에 따른 경...</td>\n",
       "      <td>23522464</td>\n",
       "      <td>아빠 육아휴직 장려금아빠 육아휴직 장려금   업무개요  남성근로자의 육아휴직을 장려...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2644</td>\n",
       "      <td>경제</td>\n",
       "      <td>[서울산업진흥원] 서울메이드란?</td>\n",
       "      <td>서울산업진흥원 서울메이드란 서울의 감성을 담은 다양하고 새로운 경험을 제공하기 위해...</td>\n",
       "      <td>23194045</td>\n",
       "      <td>[서울산업진흥원] 서울메이드란?서울산업진흥원 서울메이드란 서울의 감성을 담은 다양하...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2643</td>\n",
       "      <td>환경</td>\n",
       "      <td>(강북구) 정비중</td>\n",
       "      <td>강북구 정비중  업무개요  투명 폐트병을 교환보상하므로 수거율을 높이고 폐기물을 감...</td>\n",
       "      <td>23032485</td>\n",
       "      <td>(강북구) 정비중강북구 정비중  업무개요  투명 폐트병을 교환보상하므로 수거율을 높...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2642</td>\n",
       "      <td>복지</td>\n",
       "      <td>광진맘택시 운영(임산부,영아 양육가정 전용 택시)</td>\n",
       "      <td>광진맘택시 운영임산부영아 양육가정 전용 택시  업무개요 교통약자인 임산부와 영아가정...</td>\n",
       "      <td>22904492</td>\n",
       "      <td>광진맘택시 운영(임산부,영아 양육가정 전용 택시)광진맘택시 운영임산부영아 양육가정 ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2641</td>\n",
       "      <td>복지</td>\n",
       "      <td>마포 뇌병변장애인 비전센터</td>\n",
       "      <td>마포 뇌병변장애인 비전센터 마포뇌병변장애인 비전센터 운영   구분 내용  목적  학...</td>\n",
       "      <td>22477798</td>\n",
       "      <td>마포 뇌병변장애인 비전센터마포 뇌병변장애인 비전센터 마포뇌병변장애인 비전센터 운영 ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     번호  분류                           제목  \\\n",
       "0  2645  복지                  아빠 육아휴직 장려금   \n",
       "1  2644  경제            [서울산업진흥원] 서울메이드란?   \n",
       "2  2643  환경                    (강북구) 정비중   \n",
       "3  2642  복지  광진맘택시 운영(임산부,영아 양육가정 전용 택시)   \n",
       "4  2641  복지               마포 뇌병변장애인 비전센터   \n",
       "\n",
       "                                                  내용      내용번호  \\\n",
       "0  아빠 육아휴직 장려금   업무개요  남성근로자의 육아휴직을 장려하고 양육에 따른 경...  23522464   \n",
       "1  서울산업진흥원 서울메이드란 서울의 감성을 담은 다양하고 새로운 경험을 제공하기 위해...  23194045   \n",
       "2  강북구 정비중  업무개요  투명 폐트병을 교환보상하므로 수거율을 높이고 폐기물을 감...  23032485   \n",
       "3  광진맘택시 운영임산부영아 양육가정 전용 택시  업무개요 교통약자인 임산부와 영아가정...  22904492   \n",
       "4  마포 뇌병변장애인 비전센터 마포뇌병변장애인 비전센터 운영   구분 내용  목적  학...  22477798   \n",
       "\n",
       "                                                  문서  \n",
       "0  아빠 육아휴직 장려금아빠 육아휴직 장려금   업무개요  남성근로자의 육아휴직을 장려...  \n",
       "1  [서울산업진흥원] 서울메이드란?서울산업진흥원 서울메이드란 서울의 감성을 담은 다양하...  \n",
       "2  (강북구) 정비중강북구 정비중  업무개요  투명 폐트병을 교환보상하므로 수거율을 높...  \n",
       "3  광진맘택시 운영(임산부,영아 양육가정 전용 택시)광진맘택시 운영임산부영아 양육가정 ...  \n",
       "4  마포 뇌병변장애인 비전센터마포 뇌병변장애인 비전센터 마포뇌병변장애인 비전센터 운영 ...  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 제목과 내용을 합쳐서 문서라는 파생변수 생성\n",
    "df[\"문서\"] = df[\"제목\"] + df[\"내용\"]\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 5,
     "status": "ok",
     "timestamp": 1672570136447,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "TjuWJzWqZ0JF",
    "outputId": "a8428038-dd86-4404-fc74-d12f1d2831e0"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "행정        1098\n",
       "경제         823\n",
       "복지         217\n",
       "환경         124\n",
       "주택도시계획     110\n",
       "문화관광        96\n",
       "교통          90\n",
       "안전          51\n",
       "건강          23\n",
       "여성가족        13\n",
       "Name: 분류, dtype: int64"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# value_counts()로 분류별 빈도수 확인\n",
    "df[\"분류\"].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "executionInfo": {
     "elapsed": 4,
     "status": "ok",
     "timestamp": 1672570136448,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "0wnpc5nUZ0JF"
   },
   "outputs": [],
   "source": [
    "# 분류별 빈도수 값으로 불균형이 심해 전체 데이터로 예측을 하면 성능이 떨어질 수 있음\n",
    "# 일부 상위 분류 데이터만을 추출해 사용\n",
    "df = df[df[\"분류\"].isin([\"행정\", \"경제\", \"복지\"])]\n",
    "# df = df[df[\"분류\"].isin([\"행정\", \"경제\"])]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "executionInfo": {
     "elapsed": 6,
     "status": "ok",
     "timestamp": 1672570136946,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "zvpvDX5nZ0JG"
   },
   "outputs": [],
   "source": [
    "# 정답(label) 값\n",
    "label_name = \"분류\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "executionInfo": {
     "elapsed": 7,
     "status": "ok",
     "timestamp": 1672570136947,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "2VFsfacbZ0JG"
   },
   "outputs": [],
   "source": [
    "# 독립변수(X, 문제)와 종속변수(y, 정답)\n",
    "X = df[\"문서\"]\n",
    "y = df[label_name]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "sVldgQfrZ0JH"
   },
   "source": [
    "## label one-hot 형태로 만들기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "executionInfo": {
     "elapsed": 7,
     "status": "ok",
     "timestamp": 1672570136947,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "jcyzPconsQY5"
   },
   "outputs": [],
   "source": [
    "# get_dummies 를 사용하여 label 값을 one-hot 형태로 생성\n",
    "y_onehot = pd.get_dummies(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 6,
     "status": "ok",
     "timestamp": 1672570136947,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "JaaV4x1AsQY5",
    "outputId": "c1876702-a2af-4bf6-b3f6-f537f69a8453"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1710,), (428,), (1710, 3), (428, 3))"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# train_test_split 으로 학습과 예측에 사용할 데이터를 나누기\n",
    "# 정답값은 y_onehot 으로 지정\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(\n",
    "    X, y_onehot, test_size=0.2, random_state=42, stratify=y_onehot)\n",
    "\n",
    "X_train.shape, X_test.shape, y_train.shape, y_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 424
    },
    "executionInfo": {
     "elapsed": 5,
     "status": "ok",
     "timestamp": 1672570136947,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "3b6cgFaJsQY6",
    "outputId": "6c1294ed-a969-4d5a-d671-2618468a039f"
   },
   "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>경제</th>\n",
       "      <th>복지</th>\n",
       "      <th>행정</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2052</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2594</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1061</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>78</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>70</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1571</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1533</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>671</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>550</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1239</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1710 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      경제  복지  행정\n",
       "2052   0   0   1\n",
       "2594   1   0   0\n",
       "1061   0   0   1\n",
       "78     0   0   1\n",
       "70     0   0   1\n",
       "...   ..  ..  ..\n",
       "1571   1   0   0\n",
       "1533   1   0   0\n",
       "671    1   0   0\n",
       "550    1   0   0\n",
       "1239   0   0   1\n",
       "\n",
       "[1710 rows x 3 columns]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "경제    0.384795\n",
       "복지    0.101754\n",
       "행정    0.513450\n",
       "dtype: float64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "경제    0.385514\n",
       "복지    0.100467\n",
       "행정    0.514019\n",
       "dtype: float64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(y_train.mean())\n",
    "display(y_test.mean())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "5Vm0_ExvnAyN"
   },
   "source": [
    "## 벡터화\n",
    "### 토큰화\n",
    "\n",
    "1. 이 클래스를 사용하면 각 텍스트를 일련의 정수(각 정수는 사전에 있는 토큰의 인덱스임) 또는 단어 수에 따라 각 토큰의 계수가 이진일 수 있는 벡터로 변환하여 텍스트 말뭉치를 벡터화할 수 있습니다.(tf-idf 기반)\n",
    "\n",
    "2. 매개변수\n",
    "- num_words\n",
    ": 단어 빈도에 따라 유지할 최대 단어 수입니다. 가장 일반적인 단어 만 유지됩니다. \n",
    "-filters\n",
    ": 각 요소가 텍스트에서 필터링될 문자인 문자열입니다. 기본값은 문자를 제외한 모든 구두점과 탭 및 줄 바꿈 '입니다.\n",
    "- lower\n",
    ": 부울. 텍스트를 소문자로 변환할지 여부입니다.\n",
    "- split\n",
    ": str. 단어 분할을 위한 구분 기호입니다.\n",
    "- char_level\n",
    ": True이면 모든 문자가 토큰으로 처리됩니다.\n",
    "- oov_token\n",
    ": 주어진 경우, 그것은 word_index에 추가되고 text_to_sequence 호출 중에 어휘 밖의 단어를 대체하는 데 사용됩니다.\n",
    "\n",
    "3. 벡터화 과정\n",
    "- Tokenizer 인스턴스를 생성\n",
    "- fit_on_texts와 word_index를 사용하여 key value로 이루어진 딕셔너리를 생성\n",
    "- texts_to_sequences를 이용하여 text 문장을 숫자로 이루어진 리스트로 변경\n",
    "- 마지막으로 pad_sequences를 이용하여 리스트의 길이를 통일화\n",
    "\n",
    "\n",
    "* API Document: https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/text/Tokenizer"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "pSbB7fNT48FC"
   },
   "source": [
    "### 시퀀스 만들기"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "executionInfo": {
     "elapsed": 3064,
     "status": "ok",
     "timestamp": 1672570140007,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "hbUdn24iZ0JI",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.preprocessing.text import Tokenizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 5,
     "status": "ok",
     "timestamp": 1672570140007,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "0fkbfHCwrYbl",
    "outputId": "20056eeb-89c7-4a01-af3b-017f986a9873"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<keras.preprocessing.text.Tokenizer at 0x1360b8e80>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Tokenizer 는 데이터에 출현하는 모든 단어의 개수를 세고 빈도 수로 정렬해서 \n",
    "# num_words 에 지정된 만큼만 숫자로 반환하고, 나머지는 0 으로 반환\n",
    "# 단어 사전의 크기를 지정해 주기 위해 vocab_size를 지정\n",
    "# vocab_size는 텍스트 데이터의 전체 단어 집합의 크기\n",
    "\n",
    "vocab_size = 1000\n",
    "oov_tok = \"<oov>\"\n",
    "tokenizer = Tokenizer(num_words=vocab_size, oov_token = oov_tok)\n",
    "tokenizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "executionInfo": {
     "elapsed": 472,
     "status": "ok",
     "timestamp": 1672570140476,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "qH1gu3mIlFiD"
   },
   "outputs": [],
   "source": [
    "# Tokenizer 에 데이터 실제로 입력합니다.\n",
    "# fit_on_texts와 word_index를 사용하여 key value로 이루어진 딕셔너리를 생성\n",
    "tokenizer.fit_on_texts(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 6,
     "status": "ok",
     "timestamp": 1672570140477,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "71WWs9WVrlXD",
    "outputId": "c1a32570-257b-4c89-e8e2-cb1ff136ee15"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[\"'\",\n",
       " \"'25시\",\n",
       " \"'경복궁\",\n",
       " \"'관리주체는\",\n",
       " \"'기타운동시설운용업'의\",\n",
       " \"'노원'\",\n",
       " \"'당해\",\n",
       " \"'불허용도'란\",\n",
       " \"'새랑'의\",\n",
       " \"'성공창업\"]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# tokenizer의 word_index 속성은 단어와 숫자의 키-값 쌍을 포함하는 딕셔너리를 반환\n",
    "# 이때, 반환 시 자동으로 소문자로 변환되어 들어가며, 느낌표나 마침표 같은 구두점은 자동으로 제거\n",
    "# 각 인덱스에 해당하는 단어가 무엇인지 확인\n",
    "\n",
    "word_to_index = tokenizer.word_index\n",
    "sorted(word_to_index)[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 5,
     "status": "ok",
     "timestamp": 1672570140477,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "TSypRjUdsEB0",
    "outputId": "6409e9e3-eb42-4252-aba8-0724d8fa4b34",
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('우리아이의', 2), ('배정', 11), ('초등학교를', 2), ('알고', 45), ('싶어요', 7)]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 단어별 빈도수를 확인\n",
    "list(tokenizer.word_counts.items())[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 327
    },
    "executionInfo": {
     "elapsed": 2084,
     "status": "ok",
     "timestamp": 1672570142558,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "-481DN7osWEX",
    "outputId": "e0dc381b-f782-43cc-f6c7-7e362486d749"
   },
   "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>단어</th>\n",
       "      <th>및</th>\n",
       "      <th>돋움</th>\n",
       "      <th>수</th>\n",
       "      <th>경우</th>\n",
       "      <th>또는</th>\n",
       "      <th>등</th>\n",
       "      <th>있는</th>\n",
       "      <th>년</th>\n",
       "      <th>월</th>\n",
       "      <th>있습니다</th>\n",
       "      <th>...</th>\n",
       "      <th>궁내동</th>\n",
       "      <th>분당구</th>\n",
       "      <th>한국기기유화시험연구원</th>\n",
       "      <th>놀이시설팀</th>\n",
       "      <th>가산디지털로</th>\n",
       "      <th>제정되어</th>\n",
       "      <th>검사입니다</th>\n",
       "      <th>안전검사기관으로</th>\n",
       "      <th>시설및</th>\n",
       "      <th>설치하는가</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>빈도수</th>\n",
       "      <td>1455</td>\n",
       "      <td>1110</td>\n",
       "      <td>771</td>\n",
       "      <td>593</td>\n",
       "      <td>550</td>\n",
       "      <td>547</td>\n",
       "      <td>411</td>\n",
       "      <td>398</td>\n",
       "      <td>379</td>\n",
       "      <td>378</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1 rows × 36509 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "단어      및    돋움    수   경우   또는    등   있는    년    월  있습니다  ...  궁내동  분당구  \\\n",
       "빈도수  1455  1110  771  593  550  547  411  398  379   378  ...    1    1   \n",
       "\n",
       "단어   한국기기유화시험연구원  놀이시설팀  가산디지털로  제정되어  검사입니다  안전검사기관으로  시설및  설치하는가  \n",
       "빈도수            1      1       1     1      1         1    1      1  \n",
       "\n",
       "[1 rows x 36509 columns]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 단어별 빈도수를 확인\n",
    "word_df = pd.DataFrame(tokenizer.word_counts.items(), columns = ['단어', '빈도수']).set_index(\"단어\")\n",
    "word_df.sort_values(by=\"빈도수\", ascending=False).T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "executionInfo": {
     "elapsed": 617,
     "status": "ok",
     "timestamp": 1672570143171,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "drtrcSCkZ0JL"
   },
   "outputs": [],
   "source": [
    "# texts_to_sequences를 이용하여 text 문장을 숫자로 이루어진 리스트로 변경\n",
    "train_sequences = tokenizer.texts_to_sequences(X_train)\n",
    "test_sequences = tokenizer.texts_to_sequences(X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "MF24eCgKbBS6"
   },
   "source": [
    "## 패딩(Padding)\n",
    "\n",
    "* API Document: https://www.tensorflow.org/tutorials/keras/text_classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 6,
     "status": "ok",
     "timestamp": 1672570143172,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "cemv0J0dbF2h",
    "outputId": "9bab769f-655e-405e-84fc-9a6f48ea312d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1710,)\n"
     ]
    }
   ],
   "source": [
    "# 독립변수를 전처리합니다. \n",
    "# 문장의 길이가 제각각인 벡터의 크기를 패딩 작업을 통해 나머지 빈 공간을 0으로 채움\n",
    "# max_length는 패딩의 기준이 됨\n",
    "# padding_type='post' 는 패딩을 앞(기본값)이 아닌 뒤('post')에 채움\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "\n",
    "max_length = 500\n",
    "padding_type = \"post\"\n",
    "# padding_type = \"pre\"\n",
    "\n",
    "X_train_sp = pad_sequences(train_sequences, padding=padding_type, maxlen=max_length)\n",
    "X_test_sp = pad_sequences(test_sequences, padding=padding_type, maxlen=max_length)\n",
    "\n",
    "\n",
    "print(X_train.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "uInDm3wobb5v"
   },
   "source": [
    "## 모델 만들기\n",
    "\n",
    "### simple RNN\n",
    "\n",
    "* API Document: https://www.tensorflow.org/guide/keras/rnn\n",
    "* API Document: https://www.tensorflow.org/api_docs/python/tf/keras/layers/SimpleRNNCell"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "executionInfo": {
     "elapsed": 5,
     "status": "ok",
     "timestamp": 1672570143172,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "yUGUYDs1Z0JM"
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras import Sequential\n",
    "from tensorflow.keras.layers import Dense, Embedding, SimpleRNN, GRU, Bidirectional, LSTM, Dropout, BatchNormalization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "executionInfo": {
     "elapsed": 5,
     "status": "ok",
     "timestamp": 1672570143172,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "MbtT1Vdebei7"
   },
   "outputs": [],
   "source": [
    "# 하이퍼파라미터(모델링할 때 사용자가 직접 세팅해주는 값)을 설정\n",
    "# vocab_size는 텍스트 데이터의 전체 단어 집합의 크기\n",
    "# embedding_dim는 임베딩 할 벡터의 차원\n",
    "# max_length는 패딩의 기준\n",
    "\n",
    "embedding_dim = 64"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 4,
     "status": "ok",
     "timestamp": 1672570143172,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "ATWh1mLDZ0JN",
    "outputId": "bbbf53d1-b7a7-4e48-c474-a37917b6d629"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 클래스의 수는 분류될 예측값의 종류\n",
    "# 정답값이 one-hot 형태로 인코딩 되어 있기 때문에 정답값의 컬럼의 수가 예측값의 종류가 됨\n",
    "n_class = y_train.shape[1]\n",
    "n_class"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "V3OJM7Nv6eYu"
   },
   "source": [
    "## Bidirectional RNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "executionInfo": {
     "elapsed": 3207,
     "status": "ok",
     "timestamp": 1672570146376,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "O9RwuUKIsQY-"
   },
   "outputs": [],
   "source": [
    "model = Sequential([\n",
    "    Embedding(input_dim=vocab_size, output_dim=embedding_dim, input_length=max_length),\n",
    "    Bidirectional(LSTM(units=64, return_sequences=True)),\n",
    "    BatchNormalization(),\n",
    "    Bidirectional(LSTM(units=32)),\n",
    "    Dropout(0.2),\n",
    "    Dense(units=16, activation='relu'),\n",
    "    Dense(units=n_class, activation='softmax')\n",
    "])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "YBglNGeFZ0JN"
   },
   "source": [
    "* https://www.tensorflow.org/guide/keras/rnn#%EC%96%91%EB%B0%A9%ED%96%A5_rnn"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2bbId7zpZ0JN"
   },
   "source": [
    "### Bidirectional LSTM\n",
    "\n",
    "<img src=\"https://upload.wikimedia.org/wikipedia/commons/thumb/6/63/Long_Short-Term_Memory.svg/1600px-Long_Short-Term_Memory.svg.png\" width=\"400\">\n",
    "\n",
    "* 출처 : https://ko.wikipedia.org/wiki/%EC%88%9C%ED%99%98_%EC%8B%A0%EA%B2%BD%EB%A7%9D\n",
    "* API Document: https://www.tensorflow.org/api_docs/python/tf/keras/layers/Bidirectional"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "MJi6uCatI54I"
   },
   "source": [
    "### GRU\n",
    "\n",
    "<img src=\"https://upload.wikimedia.org/wikipedia/commons/thumb/5/5f/Gated_Recurrent_Unit.svg/1600px-Gated_Recurrent_Unit.svg.png\" width=\"400\">\n",
    "\n",
    "* 출처 : https://ko.wikipedia.org/wiki/%EC%88%9C%ED%99%98_%EC%8B%A0%EA%B2%BD%EB%A7%9D\n",
    "* API Document: https://www.tensorflow.org/api_docs/python/tf/keras/layers/GRU"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "TD6brH6LZ0JN"
   },
   "source": [
    "## 모델 컴파일"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 8,
     "status": "ok",
     "timestamp": 1672570146377,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "_uYPV_gbZ0JO",
    "outputId": "d4cd7acb-0e78-49c5-ddb7-57916cf27a32"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding (Embedding)       (None, 500, 64)           64000     \n",
      "                                                                 \n",
      " bidirectional (Bidirectiona  (None, 500, 128)         66048     \n",
      " l)                                                              \n",
      "                                                                 \n",
      " batch_normalization (BatchN  (None, 500, 128)         512       \n",
      " ormalization)                                                   \n",
      "                                                                 \n",
      " bidirectional_1 (Bidirectio  (None, 64)               41216     \n",
      " nal)                                                            \n",
      "                                                                 \n",
      " dropout (Dropout)           (None, 64)                0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 16)                1040      \n",
      "                                                                 \n",
      " dense_1 (Dense)             (None, 3)                 51        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 172,867\n",
      "Trainable params: 172,611\n",
      "Non-trainable params: 256\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# 여러개 정답 중 하나 맞추는 문제이며, 정답값이 one-hot 형태이기 때문에\n",
    "# 손실 함수는 categorical_crossentropy를 사용\n",
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer= 'adam',\n",
    "              metrics = ['accuracy']) \n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "K6Z8ZNu-Z0JO"
   },
   "source": [
    "## 학습"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "executionInfo": {
     "elapsed": 5,
     "status": "ok",
     "timestamp": 1672570146377,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "GrK3Wa0VsQZA"
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "early_stop = EarlyStopping(monitor='val_loss', patience=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 369564,
     "status": "ok",
     "timestamp": 1672570515936,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "cj9vQS6msQZA",
    "outputId": "74f969cc-c6b8-4ed5-bb4d-aceb8384d182"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "22/22 [==============================] - 33s 1s/step - loss: 0.9827 - accuracy: 0.4686 - val_loss: 1.0688 - val_accuracy: 0.4942\n",
      "Epoch 2/100\n",
      "22/22 [==============================] - 31s 1s/step - loss: 0.9358 - accuracy: 0.5190 - val_loss: 1.0559 - val_accuracy: 0.4942\n",
      "Epoch 3/100\n",
      "22/22 [==============================] - 35s 2s/step - loss: 0.9130 - accuracy: 0.5351 - val_loss: 1.0449 - val_accuracy: 0.5000\n",
      "Epoch 4/100\n",
      "22/22 [==============================] - 33s 2s/step - loss: 0.8237 - accuracy: 0.6462 - val_loss: 1.0406 - val_accuracy: 0.6170\n",
      "Epoch 5/100\n",
      "22/22 [==============================] - 32s 1s/step - loss: 0.7563 - accuracy: 0.6893 - val_loss: 0.9802 - val_accuracy: 0.5000\n",
      "Epoch 6/100\n",
      "22/22 [==============================] - 27s 1s/step - loss: 0.6903 - accuracy: 0.7208 - val_loss: 0.9599 - val_accuracy: 0.5146\n",
      "Epoch 7/100\n",
      "22/22 [==============================] - 30s 1s/step - loss: 0.5804 - accuracy: 0.7705 - val_loss: 0.9263 - val_accuracy: 0.5468\n",
      "Epoch 8/100\n",
      "22/22 [==============================] - 28s 1s/step - loss: 0.5195 - accuracy: 0.7961 - val_loss: 0.9020 - val_accuracy: 0.6053\n",
      "Epoch 9/100\n",
      "22/22 [==============================] - 27s 1s/step - loss: 0.4347 - accuracy: 0.8289 - val_loss: 0.8906 - val_accuracy: 0.6433\n",
      "Epoch 10/100\n",
      "22/22 [==============================] - 28s 1s/step - loss: 0.3791 - accuracy: 0.8640 - val_loss: 0.8410 - val_accuracy: 0.6637\n",
      "Epoch 11/100\n",
      "22/22 [==============================] - 34s 2s/step - loss: 0.3160 - accuracy: 0.8823 - val_loss: 0.8072 - val_accuracy: 0.6667\n",
      "Epoch 12/100\n",
      "22/22 [==============================] - 33s 2s/step - loss: 0.2623 - accuracy: 0.9057 - val_loss: 0.7688 - val_accuracy: 0.6608\n",
      "Epoch 13/100\n",
      "22/22 [==============================] - 30s 1s/step - loss: 0.2283 - accuracy: 0.9189 - val_loss: 0.7614 - val_accuracy: 0.6520\n",
      "Epoch 14/100\n",
      "22/22 [==============================] - 33s 1s/step - loss: 0.2031 - accuracy: 0.9254 - val_loss: 0.7826 - val_accuracy: 0.6637\n",
      "Epoch 15/100\n",
      "22/22 [==============================] - 35s 2s/step - loss: 0.1980 - accuracy: 0.9284 - val_loss: 0.8418 - val_accuracy: 0.6111\n",
      "Epoch 16/100\n",
      "22/22 [==============================] - 26s 1s/step - loss: 0.1462 - accuracy: 0.9496 - val_loss: 0.8396 - val_accuracy: 0.6462\n",
      "Epoch 17/100\n",
      "22/22 [==============================] - 31s 1s/step - loss: 0.1320 - accuracy: 0.9503 - val_loss: 0.8985 - val_accuracy: 0.6316\n",
      "Epoch 18/100\n",
      "22/22 [==============================] - 31s 1s/step - loss: 0.1237 - accuracy: 0.9613 - val_loss: 0.9586 - val_accuracy: 0.6667\n"
     ]
    }
   ],
   "source": [
    "# 모델 학습을 실행\n",
    "history = model.fit(X_train_sp, y_train, \n",
    "                    epochs=100, batch_size=64, callbacks=early_stop, validation_split=0.2,\n",
    "                    use_multiprocessing=True)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "executionInfo": {
     "elapsed": 19,
     "status": "ok",
     "timestamp": 1672570515936,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "LMnumSxuZ0JO"
   },
   "outputs": [],
   "source": [
    "# 모델 학습의 결과값을 데이터 프레임으로 만들어 확인\n",
    "df_hist = pd.DataFrame(history.history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 394
    },
    "executionInfo": {
     "elapsed": 12,
     "status": "ok",
     "timestamp": 1672570515937,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "MCU7v3_Ec5eA",
    "outputId": "0db45cb4-1c79-40cb-cb9b-ac78c67107b3"
   },
   "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>loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>val_loss</th>\n",
       "      <th>val_accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.982718</td>\n",
       "      <td>0.468567</td>\n",
       "      <td>1.068767</td>\n",
       "      <td>0.494152</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.935798</td>\n",
       "      <td>0.519006</td>\n",
       "      <td>1.055928</td>\n",
       "      <td>0.494152</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.912989</td>\n",
       "      <td>0.535088</td>\n",
       "      <td>1.044901</td>\n",
       "      <td>0.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.823704</td>\n",
       "      <td>0.646199</td>\n",
       "      <td>1.040590</td>\n",
       "      <td>0.616959</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.756251</td>\n",
       "      <td>0.689327</td>\n",
       "      <td>0.980221</td>\n",
       "      <td>0.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.690339</td>\n",
       "      <td>0.720760</td>\n",
       "      <td>0.959891</td>\n",
       "      <td>0.514620</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.580408</td>\n",
       "      <td>0.770468</td>\n",
       "      <td>0.926250</td>\n",
       "      <td>0.546784</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.519457</td>\n",
       "      <td>0.796053</td>\n",
       "      <td>0.902044</td>\n",
       "      <td>0.605263</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.434715</td>\n",
       "      <td>0.828947</td>\n",
       "      <td>0.890640</td>\n",
       "      <td>0.643275</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.379132</td>\n",
       "      <td>0.864035</td>\n",
       "      <td>0.841009</td>\n",
       "      <td>0.663743</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.315992</td>\n",
       "      <td>0.882310</td>\n",
       "      <td>0.807193</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.262327</td>\n",
       "      <td>0.905702</td>\n",
       "      <td>0.768795</td>\n",
       "      <td>0.660819</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.228263</td>\n",
       "      <td>0.918860</td>\n",
       "      <td>0.761360</td>\n",
       "      <td>0.652047</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0.203069</td>\n",
       "      <td>0.925439</td>\n",
       "      <td>0.782581</td>\n",
       "      <td>0.663743</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0.198043</td>\n",
       "      <td>0.928363</td>\n",
       "      <td>0.841750</td>\n",
       "      <td>0.611111</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.146227</td>\n",
       "      <td>0.949561</td>\n",
       "      <td>0.839597</td>\n",
       "      <td>0.646199</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.132001</td>\n",
       "      <td>0.950292</td>\n",
       "      <td>0.898541</td>\n",
       "      <td>0.631579</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>0.123685</td>\n",
       "      <td>0.961257</td>\n",
       "      <td>0.958579</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        loss  accuracy  val_loss  val_accuracy\n",
       "0   0.982718  0.468567  1.068767      0.494152\n",
       "1   0.935798  0.519006  1.055928      0.494152\n",
       "2   0.912989  0.535088  1.044901      0.500000\n",
       "3   0.823704  0.646199  1.040590      0.616959\n",
       "4   0.756251  0.689327  0.980221      0.500000\n",
       "5   0.690339  0.720760  0.959891      0.514620\n",
       "6   0.580408  0.770468  0.926250      0.546784\n",
       "7   0.519457  0.796053  0.902044      0.605263\n",
       "8   0.434715  0.828947  0.890640      0.643275\n",
       "9   0.379132  0.864035  0.841009      0.663743\n",
       "10  0.315992  0.882310  0.807193      0.666667\n",
       "11  0.262327  0.905702  0.768795      0.660819\n",
       "12  0.228263  0.918860  0.761360      0.652047\n",
       "13  0.203069  0.925439  0.782581      0.663743\n",
       "14  0.198043  0.928363  0.841750      0.611111\n",
       "15  0.146227  0.949561  0.839597      0.646199\n",
       "16  0.132001  0.950292  0.898541      0.631579\n",
       "17  0.123685  0.961257  0.958579      0.666667"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 모델 학습 결과을 그래프로 시각화\n",
    "df_hist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 283
    },
    "executionInfo": {
     "elapsed": 11,
     "status": "ok",
     "timestamp": 1672570515937,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "ixIPPgKgZ0JP",
    "outputId": "544ea466-d8b6-4a72-afa0-d5f3bd667c81"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot: >"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_hist[[\"accuracy\", \"val_accuracy\"]].plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 283
    },
    "executionInfo": {
     "elapsed": 495,
     "status": "ok",
     "timestamp": 1672570516425,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "uB6Dx0Xzty4U",
    "outputId": "0c0ad644-3cfd-4890-f59e-861f6dcd9dd9"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot: >"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_hist[[\"loss\", \"val_loss\"]].plot()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "OFpNLGzUZ0JP"
   },
   "source": [
    "## 예측"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 4729,
     "status": "ok",
     "timestamp": 1672570521151,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "o-jaBsHTt7Ri",
    "outputId": "927deaee-cebb-4676-bfc1-f3ad01d82382"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "14/14 [==============================] - 4s 136ms/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[0.01815866, 0.01040165, 0.97143966],\n",
       "       [0.08268622, 0.04059941, 0.87671435],\n",
       "       [0.87735134, 0.07421932, 0.0484294 ],\n",
       "       [0.9623527 , 0.02143038, 0.01621693],\n",
       "       [0.91504747, 0.05764544, 0.02730703],\n",
       "       [0.04691589, 0.01953024, 0.9335539 ],\n",
       "       [0.11600023, 0.0657258 , 0.818274  ],\n",
       "       [0.92846894, 0.04557508, 0.02595596],\n",
       "       [0.0122523 , 0.00732348, 0.9804243 ],\n",
       "       [0.0139066 , 0.01639604, 0.9696973 ]], dtype=float32)"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# predict() 메서드로 모델 예측\n",
    "y_pred = model.predict(X_test_sp)\n",
    "y_pred[:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "CTE70dO3Z0JP"
   },
   "source": [
    "## 평가"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 7,
     "status": "ok",
     "timestamp": 1672570521151,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "iUvRhKTVuXFR",
    "outputId": "a3e7fc58-9164-453f-8511-be61611b9119"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([2, 2, 0, 0, 0, 2, 2, 0, 2, 2])"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# numpy.argmax를 이용해 가장 큰 값의 인덱스들을 반환한 값(클래스 예측)을 y_predict에 할당\n",
    "y_predict = np.argmax(y_pred, axis=1)\n",
    "y_predict[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "executionInfo": {
     "elapsed": 4,
     "status": "ok",
     "timestamp": 1672570521152,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "WaTm4zwMZ0JP"
   },
   "outputs": [],
   "source": [
    "# numpy.argmax를 이용해 가장 큰 값의 인덱스들을 반환한 값(클래스 예측)을 y_test_val에 할당\n",
    "y_test_val = np.argmax(y_test.values, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 527,
     "status": "ok",
     "timestamp": 1672570521675,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "tXe_vRQmZ0JP",
    "outputId": "9c6fe63f-0e74-49c4-92e2-0e6b8ae38496",
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6705607476635514"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 실제값과 예측값을 비교하여 맞춘 값의 평균을 확인\n",
    "(y_test_val == y_predict).mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 4895,
     "status": "ok",
     "timestamp": 1672570526567,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "LftFDAHXIIXM",
    "outputId": "45008066-fe7d-4026-9e68-f73b63acec6d",
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "14/14 [==============================] - 3s 177ms/step - loss: 0.9565 - accuracy: 0.6706\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(0.9564558267593384, 0.6705607771873474)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 모델에 설정된 손실 값 및 메트릭 값을 반환하여 평가\n",
    "test_loss, test_acc = model.evaluate(X_test_sp, y_test)\n",
    "test_loss, test_acc "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "executionInfo": {
     "elapsed": 14,
     "status": "ok",
     "timestamp": 1672570526568,
     "user": {
      "displayName": "송영숙",
      "userId": "01138230862948206591"
     },
     "user_tz": -540
    },
    "id": "t0EVMWhcsQZB"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "provenance": []
  },
  "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": "3.10.7"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {
    "height": "calc(100% - 180px)",
    "left": "10px",
    "top": "150px",
    "width": "256px"
   },
   "toc_section_display": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}