6WEEK_BOSTON_HOUSING_DATASET-Copy1.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "05e55a8c",
   "metadata": {},
   "source": [
    "# BOSTON HOUSING DATASET\n",
    "\n",
    "목표:여러가지 데이터가 주어졌을 때 주택 가격의 중간 값을 예측해 보겠습니다.\n",
    "    \n",
    "구성:데이터 포인트가 506개로 비교적 개수가 적음. 404개는 훈련샘플, 102개는 테스트 샘플.\n",
    "    입력 데이터에 있는 각 특성은 스케일이 다름(예를들어 범죄율이 다름).\n",
    "    어떤 값은 0과 1사이의 비율을, 어떤 것은 1과 100사이의 값을 가진다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "938c66a5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(404, 13)\n",
      "(102, 13)\n"
     ]
    }
   ],
   "source": [
    "from keras.datasets import boston_housing\n",
    "\n",
    "(train_data, train_targets), (test_data, test_targets) = boston_housing.load_data() #target은 주택 가격(천달러 단위)\n",
    "\n",
    "print(train_data.shape) #404개의 훈련 샘플\n",
    "print(test_data.shape) #102개의 테스트 샘플"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "9db485ba",
   "metadata": {},
   "outputs": [],
   "source": [
    "mean = train_data.mean(axis=0) #x축의 평균(세로로 13개 값 나옴)\n",
    "train_data -= mean #각 특성에서 특성들의 평균을 뺌\n",
    "std = train_data.std(axis=0) #x축의 표준편차(세로로 13개 값 나옴)\n",
    "train_data /= std #각 평균을 뺀 값에서 특성들의 표준편차로 나눔\n",
    "\n",
    "test_data -= mean #테스트 데이터를 정규화할 때에도 훈련 데이터에서 계산한 값을 사용해야함\n",
    "test_data /= std #머신러닝에서는 테스트 데이터를 연산한 값을 사용하면 안된다"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "a4aaa8a2",
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras import models\n",
    "from keras import layers\n",
    "#훈련 데이터가 작을수록 overfitting이 쉽게 발생 -> 작은 모델을 사용하여 과대적합 방지\n",
    "def build_model():\n",
    "    model = models.Sequential()\n",
    "    model.add(layers.Dense(64,activation='relu',input_shape=(train_data.shape[1],))) #train_data.shape[1]=>13 cf.index가 0이면 404\n",
    "    model.add(layers.Dense(64,activation='relu'))\n",
    "    model.add(layers.Dense(1)) #softmax 없음->사용하게 되면 0에서1사이의 값이 나옴->집값 예측 불가능 \n",
    "    model.compile(optimizer='rmsprop',loss='mse',metrics=['mae'])\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "ae2ccf2a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "처리중인 폴드 # 0\n",
      "처리중인 폴드 # 1\n",
      "처리중인 폴드 # 2\n",
      "처리중인 폴드 # 3\n",
      "[2.0474541187286377, 2.2157647609710693, 2.3710365295410156, 2.4661128520965576]\n",
      "2.27509206533432\n"
     ]
    }
   ],
   "source": [
    "#k-fold cross-validation(k겹 교차검증)\n",
    "import numpy as np\n",
    "\n",
    "k=4 \n",
    "\n",
    "num_val_samples = len(train_data) // k\n",
    "num_epochs = 100\n",
    "all_scores = []\n",
    "for i in range(k):\n",
    "    print(\"처리중인 폴드 #\",i)\n",
    "    val_data = train_data[i*num_val_samples: (i+1) * num_val_samples] #검증데이터 준비(k번째 분할)\n",
    "    val_targets = train_targets[i*num_val_samples: (i+1) * num_val_samples]\n",
    "    \n",
    "    partial_train_data = np.concatenate( #훈련데이터 준비(다른 분할 전체)\n",
    "        [train_data[:i*num_val_samples],\n",
    "        train_data[(i+1) * num_val_samples:]],\n",
    "    axis=0)\n",
    "    partial_train_targets = np.concatenate( #concatenate=>배열 합치는 메서드\n",
    "        [train_targets[:i*num_val_samples],\n",
    "        train_targets[(i+1)*num_val_samples:]],\n",
    "    axis=0)\n",
    "    \n",
    "    model = build_model() #케라스 모델 구성(컴파일 포함) cf.위 cell에서 선언한 함수\n",
    "    model.fit(partial_train_data, partial_train_targets,\n",
    "             epochs=num_epochs, batch_size=1, verbose=0)\n",
    "    val_mse, val_mae = model.evaluate(val_data,val_targets, verbose=0) #verbose=0으로 하면 훈련 과정이 출력되지 않음\n",
    "    all_scores.append(val_mae)\n",
    "print(all_scores) #검증 세트가 다르기 때문에 검증 점수가 다르다.\n",
    "print(np.mean(all_scores)) #위의 평균값은 2.4 정도로 2 ~ 2.9까지 튀는 값들보다 신뢰도가 높음"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "4be91fc9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "처리중인 폴드 # 0\n",
      "처리중인 폴드 # 1\n",
      "처리중인 폴드 # 2\n",
      "처리중인 폴드 # 3\n"
     ]
    }
   ],
   "source": [
    "#신경망을 좀 더 오래 학습시켜본다. epochs=500만큼\n",
    "num_epochs = 500\n",
    "all_mae_historis = []\n",
    "\n",
    "for i in range(k):\n",
    "    print(\"처리중인 폴드 #\",i)\n",
    "    val_data = train_data[i*num_val_samples: (i+1) * num_val_samples] #검증데이터 준비(k번째 분할)\n",
    "    val_targets = train_targets[i*num_val_samples: (i+1) * num_val_samples]\n",
    "    \n",
    "    partial_train_data = np.concatenate( #훈련데이터 준비(다른 분할 전체)\n",
    "        [train_data[:i*num_val_samples],\n",
    "        train_data[(i+1) * num_val_samples:]],\n",
    "    axis=0)\n",
    "    partial_train_targets = np.concatenate( #concatenate=>배열 합치는 메서드\n",
    "        [train_targets[:i*num_val_samples],\n",
    "        train_targets[(i+1)*num_val_samples:]],\n",
    "    axis=0)\n",
    "    \n",
    "    model = build_model() #케라스 모델 구성(컴파일 포함) cf.위 cell에서 선언한 함수\n",
    "    history = model.fit(partial_train_data, partial_train_targets,\n",
    "              validation_data=(val_data, val_targets),\n",
    "             epochs=num_epochs, batch_size=1, verbose=0)\n",
    "    val_mse, val_mae = model.evaluate(val_data,val_targets, verbose=0) #verbose=0으로 하면 훈련 과정이 출력되지 않음\n",
    "    mae_history = history.history['val_mae']\n",
    "    all_mae_historis.append(mae_history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "a0066cf2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[4.558439373970032, 3.463166832923889, 3.1514487862586975, 3.0525707602500916, 2.8233087062835693, 2.761350393295288, 2.6971607208251953, 2.608690083026886, 2.705323815345764, 2.6141035556793213, 2.575295627117157, 2.5639345049858093, 2.4848985075950623, 2.524558186531067, 2.7657474279403687, 2.496633231639862, 2.554107367992401, 2.6235278248786926, 2.517009139060974, 2.4170278012752533, 2.4299094676971436, 2.5104681849479675, 2.4303072690963745, 2.474245250225067, 2.417906165122986, 2.5694245100021362, 2.3475985527038574, 2.5237025022506714, 2.3726430237293243, 2.4842875599861145, 2.3097556233406067, 2.381274163722992, 2.314213067293167, 2.404289960861206, 2.398387610912323, 2.3058255910873413, 2.3434820771217346, 2.332215666770935, 2.417366087436676, 2.286779284477234, 2.4192749857902527, 2.3755195438861847, 2.4668428897857666, 2.423562914133072, 2.5327998399734497, 2.3750003576278687, 2.446927696466446, 2.444058895111084, 2.4269596934318542, 2.3193222880363464, 2.3646220266819, 2.4555808305740356, 2.37227863073349, 2.531169056892395, 2.409186840057373, 2.446847140789032, 2.3379816114902496, 2.3953261971473694, 2.4076507687568665, 2.63096159696579, 2.4096254110336304, 2.3261563777923584, 2.519221842288971, 2.3477866649627686, 2.405040919780731, 2.4903813898563385, 2.4330641329288483, 2.497658908367157, 2.4831073880195618, 2.4670782685279846, 2.4688658714294434, 2.3351645469665527, 2.4980910420417786, 2.4735522270202637, 2.5709951519966125, 2.524459719657898, 2.546251595020294, 2.5696902871131897, 2.5958228707313538, 2.5178102254867554, 2.4788365364074707, 2.573840618133545, 2.4353178441524506, 2.5057194232940674, 2.437409818172455, 2.6258544325828552, 2.469709098339081, 2.4360058307647705, 2.475092828273773, 2.5927419662475586, 2.500441074371338, 2.5705307722091675, 2.4524340629577637, 2.5410566329956055, 2.418092668056488, 2.6090410351753235, 2.705419898033142, 2.5905354619026184, 2.5406380891799927, 2.6664289832115173, 2.4881551861763, 2.4589306116104126, 2.5707333087921143, 2.526416540145874, 2.616909682750702, 2.5235568284988403, 2.710036516189575, 2.509317100048065, 2.622537851333618, 2.5128705501556396, 2.550170063972473, 2.470247268676758, 2.62067312002182, 2.607655882835388, 2.753377079963684, 2.5856685042381287, 2.643766939640045, 2.6610880494117737, 2.6618855595588684, 2.570741295814514, 2.5230729579925537, 2.583919644355774, 2.6278140544891357, 2.6029319167137146, 2.7648545503616333, 2.724774420261383, 2.5941946506500244, 2.5863212943077087, 2.628045618534088, 2.6152164340019226, 2.629969298839569, 2.6199195981025696, 2.606200397014618, 2.7961291074752808, 2.5800840854644775, 2.6132872700691223, 2.6448046565055847, 2.61416357755661, 2.639782726764679, 2.6207494735717773, 2.793102741241455, 2.624042510986328, 2.691272258758545, 2.626021087169647, 2.651728928089142, 2.509339928627014, 2.7048776149749756, 2.7336679697036743, 2.7609946727752686, 2.5734872221946716, 2.6733662486076355, 2.5983537435531616, 2.711739659309387, 2.91590815782547, 2.680946111679077, 2.6268545389175415, 2.6424484252929688, 2.7773661017417908, 2.7496657371520996, 2.7653159499168396, 2.8510533571243286, 2.7103341817855835, 2.5829869508743286, 2.7926852107048035, 2.692549705505371, 2.6977267265319824, 2.716746509075165, 2.6682337522506714, 2.794472098350525, 2.8887736797332764, 2.643019676208496, 2.6393604278564453, 2.7311335802078247, 2.70681095123291, 2.7367329597473145, 2.6903937458992004, 2.672595500946045, 2.673758625984192, 2.749423325061798, 2.715299904346466, 2.7627274990081787, 2.7731606364250183, 2.7348155975341797, 2.689712166786194, 2.6876485347747803, 2.7872485518455505, 2.704841434955597, 2.81011962890625, 2.8261895775794983, 2.862282693386078, 2.7281712889671326, 2.6951542496681213, 2.7850128412246704, 2.7476394176483154, 2.6607660055160522, 2.7332810759544373, 2.717463195323944, 2.786230146884918, 2.6355289220809937, 2.687978446483612, 2.790924310684204, 2.7821938395500183, 2.874454081058502, 2.8064688444137573, 2.735628843307495, 2.700371265411377, 2.7246790528297424, 2.7902937531471252, 2.763657033443451, 2.6181498169898987, 2.724758207798004, 2.7558096051216125, 2.8814517855644226, 2.7767592072486877, 2.759797751903534, 2.6490301489830017, 2.8510526418685913, 2.71086585521698, 2.856436610221863, 2.882153272628784, 2.701029658317566, 2.674931049346924, 2.691758155822754, 2.766788363456726, 2.7323262095451355, 2.6095297932624817, 2.699421465396881, 2.8492611050605774, 2.826935350894928, 2.6358445286750793, 2.789429485797882, 2.760404407978058, 2.7145450711250305, 2.7618340849876404, 2.9427199363708496, 2.7655890583992004, 2.6973230242729187, 2.7837371826171875, 2.8374401926994324, 2.7639453411102295, 2.7629390358924866, 2.753609836101532, 2.875319719314575, 2.709738790988922, 2.727561891078949, 2.761158585548401, 2.8883041739463806, 2.800821363925934, 2.6659759283065796, 2.7623727917671204, 2.67285817861557, 2.7889971137046814, 2.753205418586731, 2.7993375062942505, 2.7424689531326294, 2.8157193660736084, 2.7098039984703064, 2.655859410762787, 2.6404044032096863, 2.638611078262329, 2.6832802295684814, 2.7259761095046997, 2.697210967540741, 2.746516227722168, 2.738270103931427, 2.787637174129486, 2.824455142021179, 2.774880111217499, 2.715131402015686, 2.705415666103363, 2.727220118045807, 2.766998589038849, 2.7862483263015747, 2.784026086330414, 2.8035308122634888, 2.849830448627472, 2.7252285480499268, 2.7647196650505066, 2.805161416530609, 2.8177490830421448, 2.9441522359848022, 2.662571609020233, 2.7202024459838867, 2.7981430888175964, 2.6886837482452393, 2.749308228492737, 2.8572874069213867, 2.6521039605140686, 2.667265295982361, 2.6979628205299377, 2.6075679063796997, 2.7643885612487793, 2.694630444049835, 2.8206663727760315, 2.686602532863617, 2.770320177078247, 2.6788788437843323, 2.730221152305603, 2.809479534626007, 2.6909435987472534, 2.7181602716445923, 2.6922128796577454, 2.761919379234314, 2.7896186113357544, 2.6546131372451782, 2.6593546867370605, 2.845602035522461, 2.852509558200836, 2.801902711391449, 2.6999934315681458, 2.7358279824256897, 2.653066873550415, 2.6611438989639282, 2.6738914847373962, 2.7440948486328125, 2.713970124721527, 2.8023598194122314, 2.737590730190277, 2.712807595729828, 2.791445255279541, 2.7224919199943542, 2.720418632030487, 2.737911105155945, 2.7156167030334473, 2.826508939266205, 2.769326150417328, 2.7808308005332947, 2.706983208656311, 2.81343537569046, 2.762689471244812, 2.7411032915115356, 2.7779557704925537, 2.7306798100471497, 2.8516699075698853, 2.7214232087135315, 2.732860028743744, 2.72428959608078, 2.8371621966362, 2.751264989376068, 2.674607276916504, 2.82831072807312, 2.8430295586586, 2.7059927582740784, 2.8337888717651367, 2.7541934847831726, 2.6818597316741943, 2.665971577167511, 2.714934766292572, 2.7294245958328247, 2.7070024013519287, 2.770634710788727, 2.653773844242096, 2.6802197694778442, 2.788007915019989, 2.6689148545265198, 2.761643171310425, 2.711484372615814, 2.645439922809601, 2.734193742275238, 2.7991663217544556, 2.6727670431137085, 2.8342255353927612, 2.7138949036598206, 2.7888245582580566, 2.693647027015686, 2.889125883579254, 2.7543317079544067, 2.7833985686302185, 2.702135741710663, 2.7718092799186707, 2.740689516067505, 2.6552165746688843, 2.6541770100593567, 2.7297788858413696, 2.7094268798828125, 2.7332613468170166, 2.7948309779167175, 2.739770531654358, 2.6937281489372253, 2.7468572854995728, 2.7154441475868225, 2.744045853614807, 2.7228471636772156, 2.8458081483840942, 2.6143072843551636, 2.768058180809021, 2.728986144065857, 2.686554968357086, 2.6896129846572876, 2.644166588783264, 2.794991970062256, 2.797752797603607, 2.704993188381195, 2.7482731342315674, 2.7073423266410828, 2.900926649570465, 2.7864899039268494, 2.7100441455841064, 2.74614155292511, 2.7712966799736023, 2.8517503142356873, 2.762357473373413, 2.7432673573493958, 2.6678508520126343, 2.81919926404953, 2.64344984292984, 2.6888781785964966, 2.7926042079925537, 2.694551646709442, 2.7540777921676636, 2.7571462988853455, 2.715936005115509, 2.721315324306488, 2.753287613391876, 2.7732518315315247, 2.7528414130210876, 2.731090009212494, 2.7695053219795227, 2.656381666660309, 2.794325590133667, 2.694640874862671, 2.710953652858734, 2.671034574508667, 2.7212976813316345, 2.712754189968109, 2.7323357462882996, 2.7651901245117188, 2.7117963433265686, 2.711615204811096, 2.6644140481948853, 2.811646342277527, 2.741801381111145, 2.7569533586502075, 2.7457816004753113, 2.667850136756897, 2.6767174005508423, 2.6767366528511047, 2.6916701197624207, 2.742424964904785, 2.8505690693855286, 2.7360341548919678, 2.7012157440185547, 2.7371163368225098, 2.6603041887283325, 2.778685986995697, 2.665480136871338, 2.712362289428711, 2.767122447490692, 2.688559889793396, 2.6521637439727783, 2.637952506542206, 2.695918560028076, 2.7781415581703186, 2.732844114303589, 2.68464195728302, 2.7266006469726562, 2.6942082047462463, 2.7645868062973022, 2.7503095865249634, 2.777502954006195, 2.8614273071289062, 2.768720269203186, 2.705879032611847, 2.8182862997055054, 2.78814435005188, 2.720534384250641, 2.8278868198394775, 2.6583462953567505, 2.74363511800766, 2.758453369140625, 2.7864677906036377, 2.767886459827423, 2.7983303666114807, 2.7593963742256165, 2.698456883430481, 2.693980634212494, 2.7640108466148376, 2.6826281547546387, 2.7213014364242554, 2.7617571353912354, 2.663272738456726, 2.6783488988876343, 2.7304237484931946, 2.7145344614982605, 2.7674614787101746, 2.6751025915145874, 2.767577052116394, 2.7731815576553345, 2.6792922616004944, 2.722886085510254, 2.687277615070343, 2.6748945116996765, 2.7605780363082886, 2.7701935172080994, 2.623776614665985, 2.6767444014549255, 2.6851927042007446, 2.6767454147338867, 2.710160732269287, 2.716050863265991]\n"
     ]
    }
   ],
   "source": [
    "#k겹 검증 점수 평균 기록\n",
    "average_mae_history = [\n",
    "    np.mean([x[i] for x in all_mae_historis]) for i in range(num_epochs)\n",
    "]\n",
    "print(average_mae_history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "8ca68bd5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(range(1,len(average_mae_history) + 1), average_mae_history)\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('validation MAE')\n",
    "plt.show() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "2c95adf8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#위 그래프는 변동이 심하여 보기 어려움\n",
    "#따라서 스케일이 많이 차이나는 첫 10개의 데이터 포인트는 제외시킨다\n",
    "#부드러운 곡선을 얻기위해 각 포인트를 이전 포인트의 지수 이동 평균으로 대체한다.\n",
    "# 지수 평균 이동은 데이터를 부드럽게 만드는 기법 중 하나로\n",
    "#이전에 계산된 이동 평균에 factor를 곱하고 현재 포인트에 (1-factor)를 곱해 합산한 것을 가리킨다.\n",
    "\n",
    "def smooth_curve(points, factor=0.9): \n",
    "    smoothed_points = []\n",
    "    for point in points: \n",
    "        if smoothed_points: # not empty\n",
    "            previous = smoothed_points[-1]\n",
    "            smoothed_points.append(previous * factor + point * (1 - factor)) \n",
    "        else:\n",
    "            smoothed_points.append(point) \n",
    "    return smoothed_points\n",
    "smooth_mae_history = smooth_curve(average_mae_history[10:]) \n",
    "plt.plot(range(1, len(smooth_mae_history) + 1), smooth_mae_history)\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Validation MAE') \n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "3c65bcc7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4/4 [==============================] - 0s 998us/step - loss: 16.9640 - mae: 2.5922\n",
      "2.592226028442383\n"
     ]
    }
   ],
   "source": [
    "model = build_model() #새롭게 컴파일된 모델을 얻는다\n",
    "model.fit(train_data, train_targets,#전체 데이터로 훈련\n",
    "         epochs=80, batch_size=16,verbose=0)\n",
    "test_mse_score, test_mae_score = model.evaluate(test_data, test_targets)\n",
    "print(test_mae_score) #2729달러 정도 차이가 남"
   ]
  }
 ],
 "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
}