12장_딥러닝_3절_8_신경망_(1)분류문제.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 개별 코드"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 랜덤 시드 설정\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "np.random.seed(0)\n",
    "tf.random.set_seed(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 데이터 불러오기\n",
    "from sklearn import datasets\n",
    "raw_wine = datasets.load_wine()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 피쳐, 타겟 데이터 지정\n",
    "X = raw_wine.data\n",
    "y = raw_wine.target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(178, 13)\n"
     ]
    }
   ],
   "source": [
    "# 피쳐 데이터 차원 확인\n",
    "print(X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{0, 1, 2}\n"
     ]
    }
   ],
   "source": [
    "# 타겟 데이터 종류 확인\n",
    "print(set(y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [1. 0. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 1. 0.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]\n",
      " [0. 0. 1.]]\n"
     ]
    }
   ],
   "source": [
    "# 타겟 데이터 원-핫 인코딩\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "y_hot = to_categorical(y) \n",
    "print(y_hot)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 트레이닝/테스트 데이터 분할\n",
    "from sklearn.model_selection import train_test_split\n",
    "X_tn, X_te, y_tn, y_te=train_test_split(X, y_hot, \n",
    "                                        random_state=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 신경망 생성\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.layers import BatchNormalization\n",
    "from tensorflow.keras.layers import Activation\n",
    "\n",
    "n_feat = X_tn.shape[1]\n",
    "n_class = len(set(y))\n",
    "epo = 30\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Dense(20, input_dim=n_feat))\n",
    "model.add(BatchNormalization())\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dense(n_class))\n",
    "model.add(Activation('softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense (Dense)                (None, 20)                280       \n",
      "_________________________________________________________________\n",
      "batch_normalization (BatchNo (None, 20)                80        \n",
      "_________________________________________________________________\n",
      "activation (Activation)      (None, 20)                0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 3)                 63        \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 3)                 0         \n",
      "=================================================================\n",
      "Total params: 423\n",
      "Trainable params: 383\n",
      "Non-trainable params: 40\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# 신경망 모형 구조 확인\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 모형 컴파일\n",
    "model.compile(loss='categorical_crossentropy', \n",
    "              optimizer='adam', \n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 133 samples\n",
      "Epoch 1/30\n",
      "133/133 [==============================] - 1s 5ms/sample - loss: 1.8539 - accuracy: 0.3835\n",
      "Epoch 2/30\n",
      "133/133 [==============================] - 0s 261us/sample - loss: 1.5996 - accuracy: 0.3985\n",
      "Epoch 3/30\n",
      "133/133 [==============================] - 0s 258us/sample - loss: 1.3056 - accuracy: 0.3534\n",
      "Epoch 4/30\n",
      "133/133 [==============================] - 0s 260us/sample - loss: 1.0861 - accuracy: 0.3910\n",
      "Epoch 5/30\n",
      "133/133 [==============================] - 0s 276us/sample - loss: 0.9303 - accuracy: 0.4737\n",
      "Epoch 6/30\n",
      "133/133 [==============================] - 0s 256us/sample - loss: 0.8378 - accuracy: 0.6241\n",
      "Epoch 7/30\n",
      "133/133 [==============================] - 0s 278us/sample - loss: 0.8117 - accuracy: 0.6241\n",
      "Epoch 8/30\n",
      "133/133 [==============================] - 0s 283us/sample - loss: 0.7908 - accuracy: 0.6090\n",
      "Epoch 9/30\n",
      "133/133 [==============================] - 0s 277us/sample - loss: 0.7725 - accuracy: 0.6165\n",
      "Epoch 10/30\n",
      "133/133 [==============================] - 0s 262us/sample - loss: 0.7599 - accuracy: 0.6316\n",
      "Epoch 11/30\n",
      "133/133 [==============================] - 0s 309us/sample - loss: 0.8250 - accuracy: 0.6316\n",
      "Epoch 12/30\n",
      "133/133 [==============================] - 0s 312us/sample - loss: 0.7003 - accuracy: 0.7143\n",
      "Epoch 13/30\n",
      "133/133 [==============================] - 0s 292us/sample - loss: 0.7944 - accuracy: 0.6992\n",
      "Epoch 14/30\n",
      "133/133 [==============================] - 0s 266us/sample - loss: 0.7255 - accuracy: 0.7669\n",
      "Epoch 15/30\n",
      "133/133 [==============================] - 0s 257us/sample - loss: 0.7025 - accuracy: 0.7218\n",
      "Epoch 16/30\n",
      "133/133 [==============================] - 0s 284us/sample - loss: 0.5666 - accuracy: 0.8346\n",
      "Epoch 17/30\n",
      "133/133 [==============================] - 0s 279us/sample - loss: 0.6220 - accuracy: 0.7820\n",
      "Epoch 18/30\n",
      "133/133 [==============================] - 0s 267us/sample - loss: 0.5833 - accuracy: 0.8271\n",
      "Epoch 19/30\n",
      "133/133 [==============================] - 0s 297us/sample - loss: 0.6032 - accuracy: 0.8120\n",
      "Epoch 20/30\n",
      "133/133 [==============================] - 0s 286us/sample - loss: 0.6131 - accuracy: 0.7669\n",
      "Epoch 21/30\n",
      "133/133 [==============================] - 0s 295us/sample - loss: 0.5810 - accuracy: 0.7970\n",
      "Epoch 22/30\n",
      "133/133 [==============================] - 0s 279us/sample - loss: 0.5826 - accuracy: 0.8496\n",
      "Epoch 23/30\n",
      "133/133 [==============================] - 0s 257us/sample - loss: 0.5616 - accuracy: 0.7895\n",
      "Epoch 24/30\n",
      "133/133 [==============================] - 0s 267us/sample - loss: 0.5340 - accuracy: 0.8271\n",
      "Epoch 25/30\n",
      "133/133 [==============================] - 0s 319us/sample - loss: 0.4504 - accuracy: 0.8496\n",
      "Epoch 26/30\n",
      "133/133 [==============================] - 0s 299us/sample - loss: 0.4844 - accuracy: 0.8346\n",
      "Epoch 27/30\n",
      "133/133 [==============================] - 0s 288us/sample - loss: 0.4063 - accuracy: 0.8722\n",
      "Epoch 28/30\n",
      "133/133 [==============================] - 0s 306us/sample - loss: 0.5498 - accuracy: 0.8120\n",
      "Epoch 29/30\n",
      "133/133 [==============================] - 0s 291us/sample - loss: 0.5067 - accuracy: 0.8195\n",
      "Epoch 30/30\n",
      "133/133 [==============================] - 0s 290us/sample - loss: 0.4318 - accuracy: 0.8571\n"
     ]
    }
   ],
   "source": [
    "# 신경망 학습\n",
    "hist = model.fit(X_tn, y_tn, epochs=epo, batch_size=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "133/1 [======================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================] - 0s 1ms/sample - loss: 0.3488 - accuracy: 0.8947\n",
      "0.8947368\n"
     ]
    }
   ],
   "source": [
    "# 트레이닝 데이터 평가\n",
    "print(model.evaluate(X_tn, y_tn)[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "45/1 [======================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================] - 0s 120us/sample - loss: 0.4090 - accuracy: 0.9333\n",
      "0.93333334\n"
     ]
    }
   ],
   "source": [
    "# 테스트 데이터 평가\n",
    "print(model.evaluate(X_te, y_te)[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24\n",
      " 25 26 27 28 29 30]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "epoch = np.arange(1,epo+1)\n",
    "print(epoch)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.38345864, 0.39849624, 0.35338345, 0.39097744, 0.47368422, 0.62406015, 0.62406015, 0.60902256, 0.6165413, 0.6315789, 0.6315789, 0.71428573, 0.69924814, 0.7669173, 0.7218045, 0.83458644, 0.7819549, 0.8270677, 0.8120301, 0.7669173, 0.7969925, 0.84962404, 0.7894737, 0.8270677, 0.84962404, 0.83458644, 0.87218046, 0.8120301, 0.81954885, 0.85714287]\n"
     ]
    }
   ],
   "source": [
    "accuracy = hist.history['accuracy'] \n",
    "print(accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1.8539354792214875, 1.5995953688047881, 1.3055668903472728, 1.086131431106338, 0.9302657313812944, 0.8378270523888725, 0.8116603114999327, 0.790762974803609, 0.7725104071143875, 0.7598636412530914, 0.8249642606964684, 0.7002793546009781, 0.794426502365815, 0.7254815018714819, 0.7024975403359062, 0.56655450893524, 0.6219969295469442, 0.5832775627311907, 0.6031937834463621, 0.6130758237121696, 0.5810363117913554, 0.5826486756927088, 0.5615835634613395, 0.5339746764279846, 0.45041022578576456, 0.4843924681943162, 0.4062664297066237, 0.5497611637850454, 0.5066769064817214, 0.4317625097762373]\n"
     ]
    }
   ],
   "source": [
    "loss = hist.history['loss'] \n",
    "print(loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "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": [
    "# 정확도 학습 그래프\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(epoch, accuracy, label='accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('accuracy')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "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": [
    "# 손실 그래프\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(epoch, loss,'r', label='loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 통합코드"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import datasets\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.layers import BatchNormalization\n",
    "from tensorflow.keras.layers import Activation\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# 랜덤 시드 설정\n",
    "np.random.seed(0)\n",
    "tf.random.set_seed(0)\n",
    "\n",
    "# 데이터 불러오기\n",
    "raw_wine = datasets.load_wine()\n",
    "\n",
    "# 피쳐, 타겟 데이터 지정\n",
    "X = raw_wine.data\n",
    "y = raw_wine.target\n",
    "\n",
    "# 피쳐 데이터 차원 확인\n",
    "print(X.shape)\n",
    "\n",
    "# 타겟 데이터 종류 확인\n",
    "print(set(y))\n",
    "\n",
    "# 타겟 데이터 원-핫 인코딩\n",
    "y_hot = to_categorical(y) \n",
    "\n",
    "# 트레이닝/테스트 데이터 분할\n",
    "X_tn, X_te, y_tn, y_te=train_test_split(X, y_hot, \n",
    "                                        random_state=0)\n",
    "\n",
    "# 신경망 생성\n",
    "n_feat = X_tn.shape[1]\n",
    "n_class = len(set(y))\n",
    "epo = 30\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Dense(20, input_dim=n_feat))\n",
    "model.add(BatchNormalization())\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dense(n_class))\n",
    "model.add(Activation('softmax'))\n",
    "\n",
    "# 신경망 모형 구조 확인\n",
    "model.summary()\n",
    "\n",
    "# 모형 컴파일\n",
    "model.compile(loss='categorical_crossentropy', \n",
    "              optimizer='adam', \n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# 신경망 학습\n",
    "hist = model.fit(X_tn, y_tn, epochs=epo, batch_size=5)\n",
    "\n",
    "# 트레이닝 데이터 평가\n",
    "print(model.evaluate(X_tn, y_tn)[1])\n",
    "\n",
    "# 테스트 데이터 평가\n",
    "print(model.evaluate(X_te, y_te)[1])\n",
    "\n",
    "\n",
    "epoch = np.arange(1,epo+1)\n",
    "accuracy = hist.history['accuracy'] \n",
    "loss = hist.history['loss'] \n",
    "\n",
    "# 정확도 학습 그래프\n",
    "plt.plot(epoch, accuracy, label='accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('accuracy')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "\n",
    "# 손실 그래프\n",
    "plt.plot(epoch, loss,'r', label='loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}