diff --git a/dataAnalysis/LAHealthcareProject/DLmodels_CNN_OLC/CNN-OLC_MultiClass.ipynb b/dataAnalysis/LAHealthcareProject/DLmodels_CNN_OLC/CNN-OLC_MultiClass.ipynb new file mode 100644 index 000000000..39e115db3 --- /dev/null +++ b/dataAnalysis/LAHealthcareProject/DLmodels_CNN_OLC/CNN-OLC_MultiClass.ipynb @@ -0,0 +1,1400 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Geospatial Analysis: Medically Underserved Areas based on Open Location Code (Multi-Class CNN-OLC Model)\n", + "\n", + "1. Data was first pre-processed and load the final version of la_facility_mua.csv here\n", + "\n", + "2. Since feature engineering has already been done, a few touch-ups before training:\n", + "\n", + "a) Feature Mapping,\n", + "\n", + "b) Dropping useless variables,\n", + "\n", + "c) Regex applications,\n", + "\n", + "d) Days Diff\n", + "\n", + "3. Data was split into Train and Validation\n", + "\n", + "4. Model: Multi-Class CNN-OLC with Tensorflow.Keras (OLC gives more granular location classification- 69 classes)\n", + "\n", + "5. The best model with Train Loss: 0.5058 - Train Accuracy: 0.7876, Val Loss: 0.4752 - Val Accuracy: 0.8150" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "os.environ['PYTHONHASHSEED']='0'" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "Using TensorFlow backend.\n" + ] + } + ], + "source": [ + "import random as rn\n", + "# Deep Learning Libraries\n", + "from tensorflow import keras\n", + "from tensorflow.keras import Sequential\n", + "from tensorflow.keras.layers import Flatten,Dense,Dropout,BatchNormalization\n", + "from tensorflow.keras.layers import Conv1D,MaxPool1D\n", + "from tensorflow.keras.optimizers import Adam\n", + "from tensorflow.keras.callbacks import EarlyStopping, ReduceLROnPlateau, ModelCheckpoint\n", + "from tensorflow.keras import callbacks\n", + "import tensorflow as tf\n", + "np.random.seed(37)\n", + "rn.seed(42)\n", + "tf.random.set_seed(42)\n", + "from keras import backend as K" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Collecting openlocationcode\r\n", + " Downloading openlocationcode-1.0.1.tar.gz (7.8 kB)\r\n", + "Building wheels for collected packages: openlocationcode\r\n", + " Building wheel for openlocationcode (setup.py) ... \u001b[?25l-\b \b\\\b \bdone\r\n", + "\u001b[?25h Created wheel for openlocationcode: filename=openlocationcode-1.0.1-py3-none-any.whl size=8419 sha256=67294f7d5e4e60a7615010ea30da6b04ec295ad67d83f7c38227cf21ea0df91f\r\n", + " Stored in directory: /root/.cache/pip/wheels/22/35/e2/e8522d642672138ffd5f7773236a5de74ceb22c99d78d9c866\r\n", + "Successfully built openlocationcode\r\n", + "Installing collected packages: openlocationcode\r\n", + "Successfully installed openlocationcode-1.0.1\r\n" + ] + } + ], + "source": [ + "!pip install openlocationcode" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "metadata": {}, + "outputs": [], + "source": [ + "#1\n", + "import pandas as pd\n", + "import numpy as np\n", + "import seaborn as sns\n", + "import re\n", + "from datetime import datetime\n", + "import matplotlib.pyplot as plt\n", + "%matplotlib inline\n", + "import matplotlib.gridspec as gridspec\n", + "import geopandas\n", + "from geopandas import GeoDataFrame\n", + "from shapely.geometry import Point\n", + "import pyproj\n", + "from openlocationcode.openlocationcode import *\n", + "from sklearn.preprocessing import StandardScaler\n", + "from sklearn.metrics import auc,roc_auc_score\n", + "from catboost import CatBoostClassifier,Pool, cv\n", + "from sklearn.preprocessing import LabelEncoder\n", + "from lightgbm import LGBMClassifier\n", + "from sklearn.metrics import roc_curve, auc, classification_report, confusion_matrix, accuracy_score\n", + "from sklearn.model_selection import train_test_split, cross_val_score, KFold, StratifiedKFold\n", + "\n", + "la_facility_mua = pd.read_csv('../input/la-final-data-2/la_facility_mua.csv')" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "metadata": {}, + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
sourcecat1cat2org_nameNameaddrln1citystatepost_iddescription...facility_address_1MUA_RuralStatusMUA_Designated_TimeMUA_Designated_DateMUA_MSSA_Name_1MUA_Updated_TimeMUA_Updated_DateMUA_PovertyMUA_ScoreMUA_Pct_100Poverty
0211Social ServicesChildren and Family Servicesneweconomicsforwomen.orgNew Economics For Women - Familysource Center ...501 S. Bixel St.Los AngelesCA56652The agency is an economic development organiza......415 South Union AvenueUrban1.067990e+122003-11-05MSSA 78.2b (Pico Union/Westlake)1.067990e+122003-11-05157757.051.00.0
1211Social ServicesChildren and Family Servicesneweconomicsforwomen.orgNew Economics For Women - Familysource Center ...501 S. Bixel St.Los AngelesCA56652The agency is an economic development organiza......600 St Paul AveUrban7.686144e+111994-05-11Westlake Service Area1.005782e+122001-11-150.060.10.0
2211Social ServicesChildren and Family Servicesneweconomicsforwomen.orgNew Economics For Women - Familysource Center ...501 S. Bixel St.Los AngelesCA56652The agency is an economic development organiza......1125 W 6th StUrban7.686144e+111994-05-11Westlake Service Area1.005782e+122001-11-150.060.10.0
3211Social ServicesChildren and Family Servicesneweconomicsforwomen.orgNew Economics For Women - Familysource Center ...501 S. Bixel St.Los AngelesCA56652The agency is an economic development organiza......1225 Wilshire BoulevardUrban7.686144e+111994-05-11Westlake Service Area1.005782e+122001-11-150.060.10.0
4211Social ServicesChildren and Family Servicesneweconomicsforwomen.orgNew Economics For Women - Familysource Center ...501 S. Bixel St.Los AngelesCA56652The agency is an economic development organiza......1127 Wilshire BlvdUrban7.686144e+111994-05-11Westlake Service Area1.005782e+122001-11-150.060.10.0
5211Social ServicesChildren and Family Servicesneweconomicsforwomen.orgNew Economics For Women - Familysource Center ...501 S. Bixel St.Los AngelesCA56652The agency is an economic development organiza......1212 Wilshire BlvdUrban7.686144e+111994-05-11Westlake Service Area1.005782e+122001-11-150.060.10.0
\n", + "

6 rows × 32 columns

\n", + "
" + ], + "text/plain": [ + " source cat1 cat2 \\\n", + "0 211 Social Services Children and Family Services \n", + "1 211 Social Services Children and Family Services \n", + "2 211 Social Services Children and Family Services \n", + "3 211 Social Services Children and Family Services \n", + "4 211 Social Services Children and Family Services \n", + "5 211 Social Services Children and Family Services \n", + "\n", + " org_name \\\n", + "0 neweconomicsforwomen.org \n", + "1 neweconomicsforwomen.org \n", + "2 neweconomicsforwomen.org \n", + "3 neweconomicsforwomen.org \n", + "4 neweconomicsforwomen.org \n", + "5 neweconomicsforwomen.org \n", + "\n", + " Name addrln1 \\\n", + "0 New Economics For Women - Familysource Center ... 501 S. Bixel St. \n", + "1 New Economics For Women - Familysource Center ... 501 S. Bixel St. \n", + "2 New Economics For Women - Familysource Center ... 501 S. Bixel St. \n", + "3 New Economics For Women - Familysource Center ... 501 S. Bixel St. \n", + "4 New Economics For Women - Familysource Center ... 501 S. Bixel St. \n", + "5 New Economics For Women - Familysource Center ... 501 S. Bixel St. \n", + "\n", + " city state post_id \\\n", + "0 Los Angeles CA 56652 \n", + "1 Los Angeles CA 56652 \n", + "2 Los Angeles CA 56652 \n", + "3 Los Angeles CA 56652 \n", + "4 Los Angeles CA 56652 \n", + "5 Los Angeles CA 56652 \n", + "\n", + " description ... \\\n", + "0 The agency is an economic development organiza... ... \n", + "1 The agency is an economic development organiza... ... \n", + "2 The agency is an economic development organiza... ... \n", + "3 The agency is an economic development organiza... ... \n", + "4 The agency is an economic development organiza... ... \n", + "5 The agency is an economic development organiza... ... \n", + "\n", + " facility_address_1 MUA_RuralStatus MUA_Designated_Time \\\n", + "0 415 South Union Avenue Urban 1.067990e+12 \n", + "1 600 St Paul Ave Urban 7.686144e+11 \n", + "2 1125 W 6th St Urban 7.686144e+11 \n", + "3 1225 Wilshire Boulevard Urban 7.686144e+11 \n", + "4 1127 Wilshire Blvd Urban 7.686144e+11 \n", + "5 1212 Wilshire Blvd Urban 7.686144e+11 \n", + "\n", + " MUA_Designated_Date MUA_MSSA_Name_1 MUA_Updated_Time \\\n", + "0 2003-11-05 MSSA 78.2b (Pico Union/Westlake) 1.067990e+12 \n", + "1 1994-05-11 Westlake Service Area 1.005782e+12 \n", + "2 1994-05-11 Westlake Service Area 1.005782e+12 \n", + "3 1994-05-11 Westlake Service Area 1.005782e+12 \n", + "4 1994-05-11 Westlake Service Area 1.005782e+12 \n", + "5 1994-05-11 Westlake Service Area 1.005782e+12 \n", + "\n", + " MUA_Updated_Date MUA_Poverty MUA_Score MUA_Pct_100Poverty \n", + "0 2003-11-05 157757.0 51.0 0.0 \n", + "1 2001-11-15 0.0 60.1 0.0 \n", + "2 2001-11-15 0.0 60.1 0.0 \n", + "3 2001-11-15 0.0 60.1 0.0 \n", + "4 2001-11-15 0.0 60.1 0.0 \n", + "5 2001-11-15 0.0 60.1 0.0 \n", + "\n", + "[6 rows x 32 columns]" + ] + }, + "execution_count": 6, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "#2\n", + "la_facility_mua.head(6)" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "metadata": {}, + "outputs": [], + "source": [ + "#3\n", + "la_facility_mua=la_facility_mua[la_facility_mua['MUA_Designated_Time'].notnull()]\n", + "la_facility_mua.drop(['facility_number_of_beds', 'facility_name', 'org_name', 'Name'],\n", + " axis=1, inplace=True)\n", + "la_facility_mua= la_facility_mua[la_facility_mua['Zip Code']!=0.0]\n", + "la_facility_mua['Zip Code']=la_facility_mua['Zip Code'].astype('int64')" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "metadata": {}, + "outputs": [], + "source": [ + "#7\n", + "la_facility_mua['facility_er_service'] = la_facility_mua['facility_er_service'].map({\n", + "'Not Applicable':'Not Applicable',\n", + "'None':'Not Applicable',\n", + "'Emergency - Basic':'Emergency',\n", + "'Emergency - Comprehensive':'Emergency',\n", + "'Emergency - Standby':'Emergency'\n", + "})" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "metadata": {}, + "outputs": [], + "source": [ + "#8\n", + "la_facility_mua['MUA_Poverty']=la_facility_mua['MUA_Poverty'].map({\n", + "157757.0:2,\n", + "129447.0:1,\n", + "0.0:0,\n", + "})" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "metadata": {}, + "outputs": [], + "source": [ + "#9\n", + "la_facility_mua['cat1+2'] = la_facility_mua['cat2'].astype('str')+'_'+la_facility_mua['cat1'].astype('str')\n", + "\n", + "la_facility_mua['city'] = la_facility_mua['city'].str.replace('Signal Hill ', 'Signal Hill')\n", + "la_facility_mua['city'] = la_facility_mua['city'].str.replace(' Hollywoo', ' Hollywood').str.replace(' Hollywoodd', ' Hollywood')\n", + "la_facility_mua['city'] = la_facility_mua['city'].str.replace('Los Angeles,', 'Los Angeles').str.replace('Los Angeles ', 'Los Angeles').str.replace('Los Angelels','Los Angeles').str.replace('Los Angeles', 'Los Angeles').str.replace(' Los Angeles', 'Los Angeles')\n", + "la_facility_mua['city'] = la_facility_mua['city'].str.replace('Industry', 'City of Industry').str.replace('City of Industry ', 'City of Industry').str.replace('City of City of Industry', 'City of Industry')\n", + "la_facility_mua['city'] = la_facility_mua['city'].str.replace('Bell', 'Bell Gardens').str.replace('Bell Gardensflower', 'Bell Gardens').str.replace('Bell Gardens Gardens', 'Bell Gardens')" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "metadata": {}, + "outputs": [], + "source": [ + "#11\n", + "la_facility_mua['Address'] = la_facility_mua['facility_address_1'].str.lower()\n", + "la_facility_mua['Street'] = la_facility_mua['Address'].str.extract('([a-zA-Z ]+)', expand=False).str.strip()\n", + "\n", + "la_facility_mua['Street'] = la_facility_mua['Street'].str.replace(' boulevard', ' blvd').str.replace(' place',' pl').str.replace(' avenue',' ave').str.replace(' street',' st').str.replace(' highway',' hwy').str.replace(' road',' rd').str.replace(' drive',' dr').str.replace(' huntington drive south', ' south huntington dr').str.replace(' huntington dr n', ' north huntington dr').str.replace('south ', 's ').str.replace('north ','n ').str.replace('east ','e ').str.replace('west ','w ').str.replace('south','s').str.replace('north','n').str.replace('east','e').str.replace('west','w')" + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "metadata": {}, + "outputs": [], + "source": [ + "#14\n", + "la_facility_mua['MUA_Designated_Date'] = pd.to_datetime(la_facility_mua['MUA_Designated_Date']) \n", + "la_facility_mua['MUA_Updated_Date'] = pd.to_datetime(la_facility_mua['MUA_Updated_Date']) \n", + "la_facility_mua['Date Difference'] = la_facility_mua['MUA_Updated_Date']-la_facility_mua['MUA_Designated_Date']" + ] + }, + { + "cell_type": "code", + "execution_count": 18, + "metadata": {}, + "outputs": [], + "source": [ + "#15\n", + "def ddiff2days(ddiff):\n", + " if not pd.isnull(ddiff):\n", + " return pd.Timedelta.total_seconds(ddiff)/(24.*3600)\n", + " else:\n", + " return np.NaN" + ] + }, + { + "cell_type": "code", + "execution_count": 19, + "metadata": {}, + "outputs": [], + "source": [ + "#16\n", + "la_facility_mua['Date Difference']=la_facility_mua['Date Difference'].apply(ddiff2days)\n", + "la_facility_mua['Date Difference']=la_facility_mua['Date Difference'].astype('int32')" + ] + }, + { + "cell_type": "code", + "execution_count": 20, + "metadata": {}, + "outputs": [], + "source": [ + "#17\n", + "la_facility_mua['design_year'] = (la_facility_mua['MUA_Designated_Date'].dt.year).astype('int32')\n", + "la_facility_mua['design_month'] = (la_facility_mua['MUA_Designated_Date'].dt.month).astype('int32')\n", + "la_facility_mua['design_week'] = (la_facility_mua['MUA_Designated_Date'].dt.week).astype('int32')\n", + "la_facility_mua['design_dow'] = (la_facility_mua['MUA_Designated_Date'].dt.dayofweek).astype('int32')\n", + "\n", + "la_facility_mua['update_year'] = (la_facility_mua['MUA_Updated_Date'].dt.year).astype('int32')\n", + "la_facility_mua['update_month'] = (la_facility_mua['MUA_Updated_Date'].dt.month).astype('int32')\n", + "la_facility_mua['update_week'] = (la_facility_mua['MUA_Updated_Date'].dt.week).astype('int32')\n", + "la_facility_mua['update_dow'] = (la_facility_mua['MUA_Updated_Date'].dt.dayofweek).astype('int32')" + ] + }, + { + "cell_type": "code", + "execution_count": 24, + "metadata": {}, + "outputs": [], + "source": [ + "#13 #69-OLC #51-ZipCode\n", + "la_facility_mua['OLC'] = la_facility_mua[[\"latitude\", \"longitude\"]].apply(lambda l: encode(l[0], l[1], codeLength=6), axis = 1)\n", + "la_facility_mua['OLC']=la_facility_mua['OLC'].astype('object')" + ] + }, + { + "cell_type": "code", + "execution_count": 33, + "metadata": {}, + "outputs": [], + "source": [ + "#24\n", + "la_facility_mua.drop([\n", + " 'MUA_Designated_Date', 'MUA_Updated_Date',\n", + " 'latitude', 'longitude', 'facility_address_1'\n", + " \n", + " ], axis=1, inplace=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 34, + "metadata": {}, + "outputs": [], + "source": [ + "cols = [c for c in la_facility_mua.columns if la_facility_mua[c].dtypes=='object']\n", + "for c in cols:\n", + " le = LabelEncoder()\n", + " la_facility_mua[c] = (le.fit_transform(la_facility_mua[c])).astype('int32')" + ] + }, + { + "cell_type": "code", + "execution_count": 35, + "metadata": {}, + "outputs": [], + "source": [ + "X = la_facility_mua.drop(labels=['OLC'], axis=1)\n", + "y = la_facility_mua['OLC'].values\n", + "\n", + "from sklearn.model_selection import train_test_split\n", + "X_train, X_cv, y_train, y_cv = train_test_split(X, y, test_size=0.25, random_state=tf.random.set_seed(42))" + ] + }, + { + "cell_type": "code", + "execution_count": 36, + "metadata": {}, + "outputs": [], + "source": [ + "X_train = X_train.to_numpy()\n", + "X_cv = X_cv.to_numpy()" + ] + }, + { + "cell_type": "code", + "execution_count": 37, + "metadata": {}, + "outputs": [], + "source": [ + "X_train = X_train.reshape(X_train.shape[0],X_train.shape[1],1)\n", + "X_cv = X_cv.reshape(X_cv.shape[0],X_cv.shape[1],1)" + ] + }, + { + "cell_type": "code", + "execution_count": 38, + "metadata": {}, + "outputs": [], + "source": [ + "# Convolutional Neural Networks (CNN)\n", + "epoch = 200\n", + "model = Sequential()\n", + "model.add(Conv1D(100,2,activation='relu',kernel_initializer='he_normal',input_shape = X_train[0].shape))\n", + "model.add(Conv1D(100,2,activation='relu', kernel_initializer='he_normal'))\n", + "model.add(Conv1D(80,2,activation='relu', kernel_initializer='he_normal'))\n", + "model.add(BatchNormalization())\n", + "model.add(Dropout(0.4))\n", + "\n", + "model.add(Flatten())\n", + "model.add(Dense(80,activation='relu', kernel_initializer='he_normal'))\n", + "model.add(Dense(69,activation='softmax'))" + ] + }, + { + "cell_type": "code", + "execution_count": 39, + "metadata": {}, + "outputs": [], + "source": [ + "model.compile(optimizer=Adam(lr=0.005), loss = 'sparse_categorical_crossentropy', \n", + " metrics=['accuracy'])" + ] + }, + { + "cell_type": "code", + "execution_count": 40, + "metadata": {}, + "outputs": [], + "source": [ + "earlystop = EarlyStopping(patience=20) \n", + "learning_rate_reduction = ReduceLROnPlateau(monitor='val_accuracy', \n", + " patience=3, \n", + " factor=0.96, \n", + " min_lr=0.00001)\n", + "checkpointer = ModelCheckpoint(filepath='CNN_model_OLC.h5', monitor='val_accuracy', verbose=1, \n", + " save_best_only=True, mode='max', period=1)\n", + "callbacks = [#earlystop, \n", + " learning_rate_reduction, checkpointer\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": 41, + "metadata": { + "scrolled": false + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Train on 7878 samples, validate on 2627 samples\n", + "Epoch 1/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 2.4234 - accuracy: 0.3562\n", + "Epoch 00001: val_accuracy improved from -inf to 0.47278, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 8s 1ms/sample - loss: 2.4165 - accuracy: 0.3569 - val_loss: 1.7498 - val_accuracy: 0.4728\n", + "Epoch 2/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 1.8138 - accuracy: 0.4454\n", + "Epoch 00002: val_accuracy improved from 0.47278 to 0.47545, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 813us/sample - loss: 1.8131 - accuracy: 0.4458 - val_loss: 1.7904 - val_accuracy: 0.4754\n", + "Epoch 3/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 1.6330 - accuracy: 0.4798\n", + "Epoch 00003: val_accuracy improved from 0.47545 to 0.53064, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 803us/sample - loss: 1.6327 - accuracy: 0.4797 - val_loss: 1.4451 - val_accuracy: 0.5306\n", + "Epoch 4/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.5392 - accuracy: 0.5045\n", + "Epoch 00004: val_accuracy did not improve from 0.53064\n", + "7878/7878 [==============================] - 7s 826us/sample - loss: 1.5388 - accuracy: 0.5044 - val_loss: 1.4463 - val_accuracy: 0.5036\n", + "Epoch 5/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 1.4870 - accuracy: 0.5085\n", + "Epoch 00005: val_accuracy did not improve from 0.53064\n", + "7878/7878 [==============================] - 7s 839us/sample - loss: 1.4889 - accuracy: 0.5084 - val_loss: 1.6329 - val_accuracy: 0.4633\n", + "Epoch 6/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.4147 - accuracy: 0.5188\n", + "Epoch 00006: val_accuracy did not improve from 0.53064\n", + "7878/7878 [==============================] - 7s 827us/sample - loss: 1.4132 - accuracy: 0.5197 - val_loss: 1.9031 - val_accuracy: 0.4949\n", + "Epoch 7/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.3741 - accuracy: 0.5403\n", + "Epoch 00007: val_accuracy did not improve from 0.53064\n", + "7878/7878 [==============================] - 6s 801us/sample - loss: 1.3751 - accuracy: 0.5399 - val_loss: 1.7566 - val_accuracy: 0.4667\n", + "Epoch 8/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 1.3071 - accuracy: 0.5570\n", + "Epoch 00008: val_accuracy did not improve from 0.53064\n", + "7878/7878 [==============================] - 6s 815us/sample - loss: 1.3086 - accuracy: 0.5565 - val_loss: 1.3109 - val_accuracy: 0.5162\n", + "Epoch 9/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 1.2804 - accuracy: 0.5571\n", + "Epoch 00009: val_accuracy improved from 0.53064 to 0.54663, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 7s 933us/sample - loss: 1.2789 - accuracy: 0.5578 - val_loss: 1.2742 - val_accuracy: 0.5466\n", + "Epoch 10/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.2479 - accuracy: 0.5703\n", + "Epoch 00010: val_accuracy improved from 0.54663 to 0.57328, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 7s 845us/sample - loss: 1.2452 - accuracy: 0.5710 - val_loss: 1.2524 - val_accuracy: 0.5733\n", + "Epoch 11/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 1.2188 - accuracy: 0.5835\n", + "Epoch 00011: val_accuracy improved from 0.57328 to 0.57823, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 809us/sample - loss: 1.2187 - accuracy: 0.5831 - val_loss: 1.2208 - val_accuracy: 0.5782\n", + "Epoch 12/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 1.1938 - accuracy: 0.5857\n", + "Epoch 00012: val_accuracy did not improve from 0.57823\n", + "7878/7878 [==============================] - 6s 802us/sample - loss: 1.1950 - accuracy: 0.5848 - val_loss: 1.5516 - val_accuracy: 0.4576\n", + "Epoch 13/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 1.1834 - accuracy: 0.5902\n", + "Epoch 00013: val_accuracy improved from 0.57823 to 0.60069, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 804us/sample - loss: 1.1840 - accuracy: 0.5901 - val_loss: 1.0681 - val_accuracy: 0.6007\n", + "Epoch 14/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.1600 - accuracy: 0.5964\n", + "Epoch 00014: val_accuracy did not improve from 0.60069\n", + "7878/7878 [==============================] - 7s 856us/sample - loss: 1.1612 - accuracy: 0.5965 - val_loss: 1.3705 - val_accuracy: 0.5219\n", + "Epoch 15/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 1.1362 - accuracy: 0.5985\n", + "Epoch 00015: val_accuracy did not improve from 0.60069\n", + "7878/7878 [==============================] - 7s 829us/sample - loss: 1.1383 - accuracy: 0.5976 - val_loss: 1.1944 - val_accuracy: 0.5744\n", + "Epoch 16/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.1240 - accuracy: 0.6037\n", + "Epoch 00016: val_accuracy improved from 0.60069 to 0.63685, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 805us/sample - loss: 1.1219 - accuracy: 0.6040 - val_loss: 1.0059 - val_accuracy: 0.6368\n", + "Epoch 17/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 1.1068 - accuracy: 0.6084\n", + "Epoch 00017: val_accuracy improved from 0.63685 to 0.64294, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 801us/sample - loss: 1.1065 - accuracy: 0.6083 - val_loss: 0.9757 - val_accuracy: 0.6429\n", + "Epoch 18/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 1.1097 - accuracy: 0.6042\n", + "Epoch 00018: val_accuracy did not improve from 0.64294\n", + "7878/7878 [==============================] - 6s 803us/sample - loss: 1.1098 - accuracy: 0.6042 - val_loss: 1.1202 - val_accuracy: 0.6045\n", + "Epoch 19/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 1.0998 - accuracy: 0.6040\n", + "Epoch 00019: val_accuracy did not improve from 0.64294\n", + "7878/7878 [==============================] - 6s 812us/sample - loss: 1.0994 - accuracy: 0.6042 - val_loss: 0.9970 - val_accuracy: 0.6292\n", + "Epoch 20/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.0951 - accuracy: 0.6094\n", + "Epoch 00020: val_accuracy did not improve from 0.64294\n", + "7878/7878 [==============================] - 6s 797us/sample - loss: 1.0951 - accuracy: 0.6092 - val_loss: 1.1139 - val_accuracy: 0.6060\n", + "Epoch 21/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.0732 - accuracy: 0.6094\n", + "Epoch 00021: val_accuracy did not improve from 0.64294\n", + "7878/7878 [==============================] - 6s 804us/sample - loss: 1.0735 - accuracy: 0.6089 - val_loss: 1.1498 - val_accuracy: 0.5836\n", + "Epoch 22/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 1.0559 - accuracy: 0.6176\n", + "Epoch 00022: val_accuracy did not improve from 0.64294\n", + "7878/7878 [==============================] - 6s 795us/sample - loss: 1.0564 - accuracy: 0.6178 - val_loss: 1.0550 - val_accuracy: 0.6079\n", + "Epoch 23/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 1.0449 - accuracy: 0.6167\n", + "Epoch 00023: val_accuracy improved from 0.64294 to 0.64598, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 7s 830us/sample - loss: 1.0449 - accuracy: 0.6168 - val_loss: 0.9769 - val_accuracy: 0.6460\n", + "Epoch 24/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.0580 - accuracy: 0.6187\n", + "Epoch 00024: val_accuracy improved from 0.64598 to 0.65131, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 7s 842us/sample - loss: 1.0607 - accuracy: 0.6177 - val_loss: 0.9688 - val_accuracy: 0.6513\n", + "Epoch 25/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 1.0274 - accuracy: 0.6282\n", + "Epoch 00025: val_accuracy did not improve from 0.65131\n", + "7878/7878 [==============================] - 6s 801us/sample - loss: 1.0277 - accuracy: 0.6285 - val_loss: 0.9469 - val_accuracy: 0.6110\n", + "Epoch 26/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 1.0342 - accuracy: 0.6256\n", + "Epoch 00026: val_accuracy did not improve from 0.65131\n", + "7878/7878 [==============================] - 6s 796us/sample - loss: 1.0348 - accuracy: 0.6255 - val_loss: 0.9499 - val_accuracy: 0.6384\n", + "Epoch 27/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 1.0147 - accuracy: 0.6297\n", + "Epoch 00027: val_accuracy did not improve from 0.65131\n", + "7878/7878 [==============================] - 6s 789us/sample - loss: 1.0144 - accuracy: 0.6297 - val_loss: 0.9641 - val_accuracy: 0.6471\n", + "Epoch 28/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.0154 - accuracy: 0.6273\n", + "Epoch 00028: val_accuracy did not improve from 0.65131\n", + "7878/7878 [==============================] - 6s 809us/sample - loss: 1.0144 - accuracy: 0.6278 - val_loss: 1.3427 - val_accuracy: 0.5158\n", + "Epoch 29/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.9947 - accuracy: 0.6368\n", + "Epoch 00029: val_accuracy improved from 0.65131 to 0.66997, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 808us/sample - loss: 0.9950 - accuracy: 0.6368 - val_loss: 0.8848 - val_accuracy: 0.6700\n", + "Epoch 30/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.9843 - accuracy: 0.6416\n", + "Epoch 00030: val_accuracy did not improve from 0.66997\n", + "7878/7878 [==============================] - 6s 791us/sample - loss: 0.9838 - accuracy: 0.6422 - val_loss: 0.9123 - val_accuracy: 0.6426\n", + "Epoch 31/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.9818 - accuracy: 0.6417\n", + "Epoch 00031: val_accuracy did not improve from 0.66997\n", + "7878/7878 [==============================] - 6s 795us/sample - loss: 0.9820 - accuracy: 0.6415 - val_loss: 0.9603 - val_accuracy: 0.6319\n", + "Epoch 32/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.9824 - accuracy: 0.6362\n", + "Epoch 00032: val_accuracy did not improve from 0.66997\n", + "7878/7878 [==============================] - 6s 802us/sample - loss: 0.9821 - accuracy: 0.6363 - val_loss: 0.9622 - val_accuracy: 0.6418\n", + "Epoch 33/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.9673 - accuracy: 0.6463\n", + "Epoch 00033: val_accuracy did not improve from 0.66997\n", + "7878/7878 [==============================] - 7s 847us/sample - loss: 0.9673 - accuracy: 0.6465 - val_loss: 0.9214 - val_accuracy: 0.6677\n", + "Epoch 34/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.9506 - accuracy: 0.6485\n", + "Epoch 00034: val_accuracy did not improve from 0.66997\n", + "7878/7878 [==============================] - 6s 808us/sample - loss: 0.9509 - accuracy: 0.6485 - val_loss: 0.8561 - val_accuracy: 0.6692\n", + "Epoch 35/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.9449 - accuracy: 0.6573\n", + "Epoch 00035: val_accuracy did not improve from 0.66997\n", + "7878/7878 [==============================] - 6s 794us/sample - loss: 0.9442 - accuracy: 0.6571 - val_loss: 1.0420 - val_accuracy: 0.6041\n", + "Epoch 36/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.9492 - accuracy: 0.6475\n", + "Epoch 00036: val_accuracy improved from 0.66997 to 0.67339, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 794us/sample - loss: 0.9497 - accuracy: 0.6475 - val_loss: 0.9035 - val_accuracy: 0.6734\n", + "Epoch 37/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.9327 - accuracy: 0.6557\n", + "Epoch 00037: val_accuracy improved from 0.67339 to 0.67568, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 794us/sample - loss: 0.9320 - accuracy: 0.6559 - val_loss: 0.8538 - val_accuracy: 0.6757\n", + "Epoch 38/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.9139 - accuracy: 0.6593\n", + "Epoch 00038: val_accuracy improved from 0.67568 to 0.68557, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 802us/sample - loss: 0.9144 - accuracy: 0.6587 - val_loss: 0.8403 - val_accuracy: 0.6856\n", + "Epoch 39/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.9193 - accuracy: 0.6565\n", + "Epoch 00039: val_accuracy did not improve from 0.68557\n", + "7878/7878 [==============================] - 6s 793us/sample - loss: 0.9182 - accuracy: 0.6569 - val_loss: 0.8422 - val_accuracy: 0.6738\n", + "Epoch 40/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.9054 - accuracy: 0.6582\n", + "Epoch 00040: val_accuracy did not improve from 0.68557\n", + "7878/7878 [==============================] - 6s 803us/sample - loss: 0.9067 - accuracy: 0.6574 - val_loss: 0.8461 - val_accuracy: 0.6856\n", + "Epoch 41/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.9114 - accuracy: 0.6611\n", + "Epoch 00041: val_accuracy improved from 0.68557 to 0.68938, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 813us/sample - loss: 0.9119 - accuracy: 0.6611 - val_loss: 0.7972 - val_accuracy: 0.6894\n", + "Epoch 42/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.9144 - accuracy: 0.6592\n", + "Epoch 00042: val_accuracy did not improve from 0.68938\n", + "7878/7878 [==============================] - 7s 887us/sample - loss: 0.9150 - accuracy: 0.6587 - val_loss: 0.8407 - val_accuracy: 0.6821\n", + "Epoch 43/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.9121 - accuracy: 0.6602\n", + "Epoch 00043: val_accuracy did not improve from 0.68938\n", + "7878/7878 [==============================] - 7s 826us/sample - loss: 0.9113 - accuracy: 0.6602 - val_loss: 0.8261 - val_accuracy: 0.6791\n", + "Epoch 44/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.8971 - accuracy: 0.6605\n", + "Epoch 00044: val_accuracy did not improve from 0.68938\n", + "7878/7878 [==============================] - 6s 800us/sample - loss: 0.8975 - accuracy: 0.6601 - val_loss: 1.1088 - val_accuracy: 0.6110\n", + "Epoch 45/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.8917 - accuracy: 0.6589\n", + "Epoch 00045: val_accuracy improved from 0.68938 to 0.69813, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 819us/sample - loss: 0.8909 - accuracy: 0.6593 - val_loss: 0.7876 - val_accuracy: 0.6981\n", + "Epoch 46/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.8936 - accuracy: 0.6644\n", + "Epoch 00046: val_accuracy improved from 0.69813 to 0.71374, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 803us/sample - loss: 0.8918 - accuracy: 0.6651 - val_loss: 0.7844 - val_accuracy: 0.7137\n", + "Epoch 47/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.8807 - accuracy: 0.6700\n", + "Epoch 00047: val_accuracy did not improve from 0.71374\n", + "7878/7878 [==============================] - 6s 811us/sample - loss: 0.8801 - accuracy: 0.6702 - val_loss: 1.4481 - val_accuracy: 0.5383\n", + "Epoch 48/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.8678 - accuracy: 0.6691\n", + "Epoch 00048: val_accuracy did not improve from 0.71374\n", + "7878/7878 [==============================] - 6s 786us/sample - loss: 0.8676 - accuracy: 0.6688 - val_loss: 0.8374 - val_accuracy: 0.6829\n", + "Epoch 49/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.8706 - accuracy: 0.6709\n", + "Epoch 00049: val_accuracy did not improve from 0.71374\n", + "7878/7878 [==============================] - 6s 789us/sample - loss: 0.8723 - accuracy: 0.6702 - val_loss: 1.0226 - val_accuracy: 0.6258\n", + "Epoch 50/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.8792 - accuracy: 0.6672\n", + "Epoch 00050: val_accuracy did not improve from 0.71374\n", + "7878/7878 [==============================] - 6s 805us/sample - loss: 0.8790 - accuracy: 0.6670 - val_loss: 0.8635 - val_accuracy: 0.6608\n", + "Epoch 51/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.8686 - accuracy: 0.6774\n", + "Epoch 00051: val_accuracy did not improve from 0.71374\n", + "7878/7878 [==============================] - 6s 804us/sample - loss: 0.8682 - accuracy: 0.6775 - val_loss: 0.7701 - val_accuracy: 0.7073\n", + "Epoch 52/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.8619 - accuracy: 0.6733\n", + "Epoch 00052: val_accuracy did not improve from 0.71374\n", + "7878/7878 [==============================] - 7s 831us/sample - loss: 0.8612 - accuracy: 0.6739 - val_loss: 0.7697 - val_accuracy: 0.7137\n", + "Epoch 53/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.8632 - accuracy: 0.6733\n", + "Epoch 00053: val_accuracy did not improve from 0.71374\n", + "7878/7878 [==============================] - 6s 799us/sample - loss: 0.8627 - accuracy: 0.6729 - val_loss: 0.8079 - val_accuracy: 0.6978\n", + "Epoch 54/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.8506 - accuracy: 0.6793\n", + "Epoch 00054: val_accuracy did not improve from 0.71374\n", + "7878/7878 [==============================] - 6s 795us/sample - loss: 0.8507 - accuracy: 0.6791 - val_loss: 0.9189 - val_accuracy: 0.6433\n", + "Epoch 55/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.8316 - accuracy: 0.6841\n", + "Epoch 00055: val_accuracy did not improve from 0.71374\n", + "7878/7878 [==============================] - 6s 795us/sample - loss: 0.8324 - accuracy: 0.6838 - val_loss: 0.7667 - val_accuracy: 0.6901\n", + "Epoch 56/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.8405 - accuracy: 0.6801\n", + "Epoch 00057: val_accuracy did not improve from 0.71374\n", + "7878/7878 [==============================] - 7s 839us/sample - loss: 0.8404 - accuracy: 0.6800 - val_loss: 0.7386 - val_accuracy: 0.6978\n", + "Epoch 58/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.8136 - accuracy: 0.6811\n", + "Epoch 00058: val_accuracy did not improve from 0.71374\n", + "7878/7878 [==============================] - 7s 902us/sample - loss: 0.8143 - accuracy: 0.6816 - val_loss: 1.2225 - val_accuracy: 0.5866\n", + "Epoch 59/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.8181 - accuracy: 0.6810\n", + "Epoch 00059: val_accuracy improved from 0.71374 to 0.72745, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 814us/sample - loss: 0.8178 - accuracy: 0.6811 - val_loss: 0.7134 - val_accuracy: 0.7274\n", + "Epoch 60/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.8183 - accuracy: 0.6936\n", + "Epoch 00060: val_accuracy did not improve from 0.72745\n", + "7878/7878 [==============================] - 6s 803us/sample - loss: 0.8172 - accuracy: 0.6935 - val_loss: 0.7498 - val_accuracy: 0.7122\n", + "Epoch 61/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.8199 - accuracy: 0.6783\n", + "Epoch 00061: val_accuracy did not improve from 0.72745\n", + "7878/7878 [==============================] - 7s 846us/sample - loss: 0.8189 - accuracy: 0.6790 - val_loss: 0.7478 - val_accuracy: 0.7031\n", + "Epoch 62/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.8121 - accuracy: 0.6897\n", + "Epoch 00062: val_accuracy did not improve from 0.72745\n", + "7878/7878 [==============================] - 6s 817us/sample - loss: 0.8119 - accuracy: 0.6898 - val_loss: 0.7464 - val_accuracy: 0.7073\n", + "Epoch 63/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.8061 - accuracy: 0.6814\n", + "Epoch 00063: val_accuracy did not improve from 0.72745\n", + "7878/7878 [==============================] - 6s 786us/sample - loss: 0.8054 - accuracy: 0.6815 - val_loss: 0.7178 - val_accuracy: 0.7214\n", + "Epoch 64/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.7954 - accuracy: 0.6880\n", + "Epoch 00064: val_accuracy did not improve from 0.72745\n", + "7878/7878 [==============================] - 6s 798us/sample - loss: 0.7969 - accuracy: 0.6874 - val_loss: 0.7031 - val_accuracy: 0.7267\n", + "Epoch 65/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.7987 - accuracy: 0.6922\n", + "Epoch 00065: val_accuracy did not improve from 0.72745\n", + "7878/7878 [==============================] - 6s 793us/sample - loss: 0.7983 - accuracy: 0.6927 - val_loss: 0.7708 - val_accuracy: 0.6970\n", + "Epoch 66/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.8006 - accuracy: 0.6931\n", + "Epoch 00066: val_accuracy did not improve from 0.72745\n", + "7878/7878 [==============================] - 6s 797us/sample - loss: 0.8006 - accuracy: 0.6929 - val_loss: 0.7635 - val_accuracy: 0.7069\n", + "Epoch 67/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.7823 - accuracy: 0.6942\n", + "Epoch 00067: val_accuracy did not improve from 0.72745\n", + "7878/7878 [==============================] - 6s 788us/sample - loss: 0.7814 - accuracy: 0.6948 - val_loss: 0.7202 - val_accuracy: 0.7202\n", + "Epoch 68/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.7656 - accuracy: 0.7033\n", + "Epoch 00068: val_accuracy did not improve from 0.72745\n", + "7878/7878 [==============================] - 6s 786us/sample - loss: 0.7660 - accuracy: 0.7035 - val_loss: 0.7375 - val_accuracy: 0.6920\n", + "Epoch 69/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.7741 - accuracy: 0.7003\n", + "Epoch 00069: val_accuracy did not improve from 0.72745\n", + "7878/7878 [==============================] - 6s 791us/sample - loss: 0.7739 - accuracy: 0.7003 - val_loss: 0.7619 - val_accuracy: 0.7099\n", + "Epoch 70/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.7637 - accuracy: 0.7057\n", + "Epoch 00070: val_accuracy did not improve from 0.72745\n", + "7878/7878 [==============================] - 7s 849us/sample - loss: 0.7646 - accuracy: 0.7059 - val_loss: 0.6874 - val_accuracy: 0.7172\n", + "Epoch 71/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.7746 - accuracy: 0.7006\n", + "Epoch 00071: val_accuracy did not improve from 0.72745\n", + "7878/7878 [==============================] - 6s 820us/sample - loss: 0.7727 - accuracy: 0.7017 - val_loss: 0.6801 - val_accuracy: 0.7233\n", + "Epoch 72/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.7648 - accuracy: 0.6995\n", + "Epoch 00072: val_accuracy improved from 0.72745 to 0.73848, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 806us/sample - loss: 0.7654 - accuracy: 0.6994 - val_loss: 0.7058 - val_accuracy: 0.7385\n", + "Epoch 73/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.7613 - accuracy: 0.7064\n", + "Epoch 00073: val_accuracy improved from 0.73848 to 0.75105, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 805us/sample - loss: 0.7614 - accuracy: 0.7064 - val_loss: 0.6647 - val_accuracy: 0.7510\n", + "Epoch 74/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.7485 - accuracy: 0.7076\n", + "Epoch 00074: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 792us/sample - loss: 0.7483 - accuracy: 0.7078 - val_loss: 0.7979 - val_accuracy: 0.7050\n", + "Epoch 75/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.7671 - accuracy: 0.7012\n", + "Epoch 00075: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 788us/sample - loss: 0.7667 - accuracy: 0.7014 - val_loss: 0.7245 - val_accuracy: 0.7305\n", + "Epoch 76/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.7468 - accuracy: 0.7077\n", + "Epoch 00076: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 808us/sample - loss: 0.7479 - accuracy: 0.7077 - val_loss: 0.6872 - val_accuracy: 0.7301\n", + "Epoch 77/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.7465 - accuracy: 0.7054\n", + "Epoch 00077: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 786us/sample - loss: 0.7470 - accuracy: 0.7051 - val_loss: 0.7096 - val_accuracy: 0.7195\n", + "Epoch 78/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.7397 - accuracy: 0.7083\n", + "Epoch 00078: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 793us/sample - loss: 0.7401 - accuracy: 0.7082 - val_loss: 0.7170 - val_accuracy: 0.7301\n", + "Epoch 79/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.7340 - accuracy: 0.7139\n", + "Epoch 00079: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 782us/sample - loss: 0.7342 - accuracy: 0.7139 - val_loss: 0.6848 - val_accuracy: 0.7229\n", + "Epoch 80/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.7303 - accuracy: 0.7143\n", + "Epoch 00080: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 815us/sample - loss: 0.7298 - accuracy: 0.7144 - val_loss: 0.8567 - val_accuracy: 0.6650\n", + "Epoch 81/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.7388 - accuracy: 0.7048\n", + "Epoch 00081: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 7s 829us/sample - loss: 0.7380 - accuracy: 0.7045 - val_loss: 0.6975 - val_accuracy: 0.7290\n", + "Epoch 82/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.7237 - accuracy: 0.7130\n", + "Epoch 00082: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 782us/sample - loss: 0.7234 - accuracy: 0.7133 - val_loss: 0.6471 - val_accuracy: 0.7392\n", + "Epoch 83/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.7221 - accuracy: 0.7151\n", + "Epoch 00083: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 785us/sample - loss: 0.7230 - accuracy: 0.7146 - val_loss: 0.6721 - val_accuracy: 0.7328\n", + "Epoch 84/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.7286 - accuracy: 0.7052\n", + "Epoch 00084: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 784us/sample - loss: 0.7287 - accuracy: 0.7050 - val_loss: 0.6571 - val_accuracy: 0.7484\n", + "Epoch 85/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.7186 - accuracy: 0.7166\n", + "Epoch 00085: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 790us/sample - loss: 0.7186 - accuracy: 0.7169 - val_loss: 0.6461 - val_accuracy: 0.7389\n", + "Epoch 86/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.7129 - accuracy: 0.7221\n", + "Epoch 00086: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 790us/sample - loss: 0.7127 - accuracy: 0.7221 - val_loss: 0.6849 - val_accuracy: 0.7229\n", + "Epoch 87/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.7066 - accuracy: 0.7170\n", + "Epoch 00087: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 790us/sample - loss: 0.7086 - accuracy: 0.7163 - val_loss: 0.6342 - val_accuracy: 0.7503\n", + "Epoch 88/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.7324 - accuracy: 0.7119\n", + "Epoch 00088: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 786us/sample - loss: 0.7322 - accuracy: 0.7120 - val_loss: 0.7300 - val_accuracy: 0.7096\n", + "Epoch 89/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6868 - accuracy: 0.7260\n", + "Epoch 00089: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 811us/sample - loss: 0.6867 - accuracy: 0.7259 - val_loss: 0.6435 - val_accuracy: 0.7354\n", + "Epoch 90/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6984 - accuracy: 0.7237\n", + "Epoch 00090: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 7s 873us/sample - loss: 0.6982 - accuracy: 0.7238 - val_loss: 0.6346 - val_accuracy: 0.7472\n", + "Epoch 91/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6901 - accuracy: 0.7273\n", + "Epoch 00091: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 801us/sample - loss: 0.6901 - accuracy: 0.7273 - val_loss: 0.6982 - val_accuracy: 0.7313\n", + "Epoch 92/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6936 - accuracy: 0.7257\n", + "Epoch 00092: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 788us/sample - loss: 0.6937 - accuracy: 0.7257 - val_loss: 0.6519 - val_accuracy: 0.7431\n", + "Epoch 93/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6815 - accuracy: 0.7298\n", + "Epoch 00093: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 779us/sample - loss: 0.6823 - accuracy: 0.7291 - val_loss: 0.6476 - val_accuracy: 0.7408\n", + "Epoch 94/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6720 - accuracy: 0.7263\n", + "Epoch 00094: val_accuracy did not improve from 0.75105\n", + "7878/7878 [==============================] - 6s 788us/sample - loss: 0.6716 - accuracy: 0.7266 - val_loss: 0.7119 - val_accuracy: 0.7271\n", + "Epoch 95/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6712 - accuracy: 0.7295\n", + "Epoch 00095: val_accuracy improved from 0.75105 to 0.75447, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 791us/sample - loss: 0.6699 - accuracy: 0.7295 - val_loss: 0.6186 - val_accuracy: 0.7545\n", + "Epoch 96/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6770 - accuracy: 0.7269\n", + "Epoch 00096: val_accuracy did not improve from 0.75447\n", + "7878/7878 [==============================] - 6s 784us/sample - loss: 0.6779 - accuracy: 0.7268 - val_loss: 0.6469 - val_accuracy: 0.7362\n", + "Epoch 97/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6815 - accuracy: 0.7258\n", + "Epoch 00097: val_accuracy did not improve from 0.75447\n", + "7878/7878 [==============================] - 6s 807us/sample - loss: 0.6799 - accuracy: 0.7265 - val_loss: 0.6345 - val_accuracy: 0.7484\n", + "Epoch 98/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6577 - accuracy: 0.7326\n", + "Epoch 00098: val_accuracy did not improve from 0.75447\n", + "7878/7878 [==============================] - 6s 782us/sample - loss: 0.6578 - accuracy: 0.7327 - val_loss: 0.6713 - val_accuracy: 0.7453\n", + "Epoch 99/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6749 - accuracy: 0.7308\n", + "Epoch 00099: val_accuracy improved from 0.75447 to 0.76209, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 817us/sample - loss: 0.6745 - accuracy: 0.7306 - val_loss: 0.6044 - val_accuracy: 0.7621\n", + "Epoch 100/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6680 - accuracy: 0.7296\n", + "Epoch 00100: val_accuracy did not improve from 0.76209\n", + "7878/7878 [==============================] - 6s 812us/sample - loss: 0.6676 - accuracy: 0.7298 - val_loss: 0.6225 - val_accuracy: 0.7499\n", + "Epoch 101/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6587 - accuracy: 0.7344\n", + "Epoch 00101: val_accuracy did not improve from 0.76209\n", + "7878/7878 [==============================] - 6s 788us/sample - loss: 0.6590 - accuracy: 0.7343 - val_loss: 0.6477 - val_accuracy: 0.7301\n", + "Epoch 102/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6629 - accuracy: 0.7301\n", + "Epoch 00102: val_accuracy did not improve from 0.76209\n", + "7878/7878 [==============================] - 6s 793us/sample - loss: 0.6630 - accuracy: 0.7299 - val_loss: 0.6203 - val_accuracy: 0.7461\n", + "Epoch 103/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6539 - accuracy: 0.7332\n", + "Epoch 00103: val_accuracy did not improve from 0.76209\n", + "7878/7878 [==============================] - 6s 785us/sample - loss: 0.6536 - accuracy: 0.7333 - val_loss: 0.6575 - val_accuracy: 0.7499\n", + "Epoch 104/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6535 - accuracy: 0.7335\n", + "Epoch 00104: val_accuracy did not improve from 0.76209\n", + "7878/7878 [==============================] - 6s 789us/sample - loss: 0.6539 - accuracy: 0.7337 - val_loss: 0.9147 - val_accuracy: 0.6875\n", + "Epoch 105/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6575 - accuracy: 0.7379\n", + "Epoch 00105: val_accuracy did not improve from 0.76209\n", + "7878/7878 [==============================] - 6s 785us/sample - loss: 0.6576 - accuracy: 0.7379 - val_loss: 0.6607 - val_accuracy: 0.7309\n", + "Epoch 106/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6387 - accuracy: 0.7405\n", + "Epoch 00106: val_accuracy did not improve from 0.76209\n", + "7878/7878 [==============================] - 6s 815us/sample - loss: 0.6383 - accuracy: 0.7407 - val_loss: 0.6372 - val_accuracy: 0.7533\n", + "Epoch 107/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6531 - accuracy: 0.7380\n", + "Epoch 00107: val_accuracy improved from 0.76209 to 0.77008, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 7s 900us/sample - loss: 0.6526 - accuracy: 0.7383 - val_loss: 0.5885 - val_accuracy: 0.7701\n", + "Epoch 108/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6469 - accuracy: 0.7407\n", + "Epoch 00108: val_accuracy did not improve from 0.77008\n", + "7878/7878 [==============================] - 7s 884us/sample - loss: 0.6468 - accuracy: 0.7407 - val_loss: 0.5827 - val_accuracy: 0.7636\n", + "Epoch 109/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6486 - accuracy: 0.7360\n", + "Epoch 00109: val_accuracy did not improve from 0.77008\n", + "7878/7878 [==============================] - 6s 818us/sample - loss: 0.6493 - accuracy: 0.7358 - val_loss: 0.6038 - val_accuracy: 0.7625\n", + "Epoch 110/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6369 - accuracy: 0.7445\n", + "Epoch 00110: val_accuracy did not improve from 0.77008\n", + "7878/7878 [==============================] - 6s 795us/sample - loss: 0.6377 - accuracy: 0.7441 - val_loss: 0.5968 - val_accuracy: 0.7609\n", + "Epoch 111/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6357 - accuracy: 0.7414\n", + "Epoch 00111: val_accuracy improved from 0.77008 to 0.77046, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 798us/sample - loss: 0.6354 - accuracy: 0.7416 - val_loss: 0.5922 - val_accuracy: 0.7705\n", + "Epoch 112/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6286 - accuracy: 0.7443\n", + "Epoch 00112: val_accuracy did not improve from 0.77046\n", + "7878/7878 [==============================] - 6s 794us/sample - loss: 0.6282 - accuracy: 0.7444 - val_loss: 0.5878 - val_accuracy: 0.7674\n", + "Epoch 113/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6298 - accuracy: 0.7425\n", + "Epoch 00113: val_accuracy improved from 0.77046 to 0.77845, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 807us/sample - loss: 0.6290 - accuracy: 0.7427 - val_loss: 0.5674 - val_accuracy: 0.7785\n", + "Epoch 114/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6264 - accuracy: 0.7504\n", + "Epoch 00114: val_accuracy did not improve from 0.77845\n", + "7878/7878 [==============================] - 6s 781us/sample - loss: 0.6281 - accuracy: 0.7496 - val_loss: 0.6326 - val_accuracy: 0.7587\n", + "Epoch 115/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6244 - accuracy: 0.7466\n", + "Epoch 00115: val_accuracy improved from 0.77845 to 0.78112, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 792us/sample - loss: 0.6246 - accuracy: 0.7464 - val_loss: 0.5795 - val_accuracy: 0.7811\n", + "Epoch 116/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6257 - accuracy: 0.7439\n", + "Epoch 00116: val_accuracy did not improve from 0.78112\n", + "7878/7878 [==============================] - 6s 788us/sample - loss: 0.6246 - accuracy: 0.7446 - val_loss: 0.5695 - val_accuracy: 0.7762\n", + "Epoch 117/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6198 - accuracy: 0.7488\n", + "Epoch 00117: val_accuracy did not improve from 0.78112\n", + "7878/7878 [==============================] - 6s 785us/sample - loss: 0.6195 - accuracy: 0.7485 - val_loss: 0.5652 - val_accuracy: 0.7777\n", + "Epoch 118/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6230 - accuracy: 0.7492\n", + "Epoch 00118: val_accuracy improved from 0.78112 to 0.78188, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 7s 848us/sample - loss: 0.6248 - accuracy: 0.7483 - val_loss: 0.5587 - val_accuracy: 0.7819\n", + "Epoch 119/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6180 - accuracy: 0.7480\n", + "Epoch 00119: val_accuracy did not improve from 0.78188\n", + "7878/7878 [==============================] - 6s 794us/sample - loss: 0.6180 - accuracy: 0.7480 - val_loss: 0.5683 - val_accuracy: 0.7769\n", + "Epoch 120/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6220 - accuracy: 0.7423\n", + "Epoch 00120: val_accuracy did not improve from 0.78188\n", + "7878/7878 [==============================] - 6s 789us/sample - loss: 0.6221 - accuracy: 0.7424 - val_loss: 0.5597 - val_accuracy: 0.7781\n", + "Epoch 121/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6203 - accuracy: 0.7473\n", + "Epoch 00121: val_accuracy did not improve from 0.78188\n", + "7878/7878 [==============================] - 6s 785us/sample - loss: 0.6201 - accuracy: 0.7471 - val_loss: 0.5685 - val_accuracy: 0.7750\n", + "Epoch 122/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6084 - accuracy: 0.7481\n", + "Epoch 00122: val_accuracy did not improve from 0.78188\n", + "7878/7878 [==============================] - 6s 790us/sample - loss: 0.6088 - accuracy: 0.7480 - val_loss: 0.5581 - val_accuracy: 0.7739\n", + "Epoch 123/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6054 - accuracy: 0.7524\n", + "Epoch 00123: val_accuracy improved from 0.78188 to 0.78797, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 809us/sample - loss: 0.6053 - accuracy: 0.7525 - val_loss: 0.5481 - val_accuracy: 0.7880\n", + "Epoch 124/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6015 - accuracy: 0.7510\n", + "Epoch 00124: val_accuracy did not improve from 0.78797\n", + "7878/7878 [==============================] - 6s 783us/sample - loss: 0.6011 - accuracy: 0.7511 - val_loss: 0.5504 - val_accuracy: 0.7830\n", + "Epoch 125/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.6060 - accuracy: 0.7548\n", + "Epoch 00125: val_accuracy did not improve from 0.78797\n", + "7878/7878 [==============================] - 6s 789us/sample - loss: 0.6062 - accuracy: 0.7550 - val_loss: 0.5549 - val_accuracy: 0.7766\n", + "Epoch 126/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5959 - accuracy: 0.7572\n", + "Epoch 00126: val_accuracy did not improve from 0.78797\n", + "7878/7878 [==============================] - 6s 789us/sample - loss: 0.5956 - accuracy: 0.7573 - val_loss: 0.5881 - val_accuracy: 0.7674\n", + "Epoch 127/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.6098 - accuracy: 0.7551\n", + "Epoch 00127: val_accuracy did not improve from 0.78797\n", + "7878/7878 [==============================] - 6s 817us/sample - loss: 0.6097 - accuracy: 0.7551 - val_loss: 0.5684 - val_accuracy: 0.7838\n", + "Epoch 128/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5894 - accuracy: 0.7591\n", + "Epoch 00128: val_accuracy improved from 0.78797 to 0.78873, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 7s 838us/sample - loss: 0.5892 - accuracy: 0.7592 - val_loss: 0.5295 - val_accuracy: 0.7887\n", + "Epoch 129/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5824 - accuracy: 0.7617\n", + "Epoch 00129: val_accuracy improved from 0.78873 to 0.79254, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 817us/sample - loss: 0.5827 - accuracy: 0.7614 - val_loss: 0.5358 - val_accuracy: 0.7925\n", + "Epoch 130/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.5846 - accuracy: 0.7633\n", + "Epoch 00130: val_accuracy did not improve from 0.79254\n", + "7878/7878 [==============================] - 6s 811us/sample - loss: 0.5857 - accuracy: 0.7629 - val_loss: 0.5592 - val_accuracy: 0.7884\n", + "Epoch 131/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.5927 - accuracy: 0.7584\n", + "Epoch 00131: val_accuracy did not improve from 0.79254\n", + "7878/7878 [==============================] - 6s 805us/sample - loss: 0.5933 - accuracy: 0.7578 - val_loss: 0.5485 - val_accuracy: 0.7895\n", + "Epoch 132/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5875 - accuracy: 0.7570\n", + "Epoch 00132: val_accuracy improved from 0.79254 to 0.79368, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 825us/sample - loss: 0.5874 - accuracy: 0.7570 - val_loss: 0.5331 - val_accuracy: 0.7937\n", + "Epoch 133/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5721 - accuracy: 0.7641\n", + "Epoch 00133: val_accuracy did not improve from 0.79368\n", + "7878/7878 [==============================] - 6s 819us/sample - loss: 0.5722 - accuracy: 0.7640 - val_loss: 0.5747 - val_accuracy: 0.7785\n", + "Epoch 134/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5805 - accuracy: 0.7599\n", + "Epoch 00134: val_accuracy improved from 0.79368 to 0.79482, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 819us/sample - loss: 0.5802 - accuracy: 0.7601 - val_loss: 0.5341 - val_accuracy: 0.7948\n", + "Epoch 135/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5866 - accuracy: 0.7608\n", + "Epoch 00135: val_accuracy did not improve from 0.79482\n", + "7878/7878 [==============================] - 6s 817us/sample - loss: 0.5865 - accuracy: 0.7609 - val_loss: 0.5295 - val_accuracy: 0.7910\n", + "Epoch 136/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5768 - accuracy: 0.7634\n", + "Epoch 00136: val_accuracy did not improve from 0.79482\n", + "7878/7878 [==============================] - 6s 807us/sample - loss: 0.5768 - accuracy: 0.7638 - val_loss: 0.5291 - val_accuracy: 0.7918\n", + "Epoch 137/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5734 - accuracy: 0.7673\n", + "Epoch 00137: val_accuracy did not improve from 0.79482\n", + "7878/7878 [==============================] - 7s 852us/sample - loss: 0.5734 - accuracy: 0.7673 - val_loss: 0.5127 - val_accuracy: 0.7944\n", + "Epoch 138/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.5730 - accuracy: 0.7642\n", + "Epoch 00138: val_accuracy did not improve from 0.79482\n", + "7878/7878 [==============================] - 7s 895us/sample - loss: 0.5731 - accuracy: 0.7642 - val_loss: 0.5567 - val_accuracy: 0.7899\n", + "Epoch 139/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5572 - accuracy: 0.7767\n", + "Epoch 00139: val_accuracy improved from 0.79482 to 0.80206, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 813us/sample - loss: 0.5572 - accuracy: 0.7767 - val_loss: 0.5056 - val_accuracy: 0.8021\n", + "Epoch 140/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.5686 - accuracy: 0.7667\n", + "Epoch 00140: val_accuracy did not improve from 0.80206\n", + "7878/7878 [==============================] - 6s 809us/sample - loss: 0.5687 - accuracy: 0.7667 - val_loss: 0.5158 - val_accuracy: 0.7891\n", + "Epoch 141/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5687 - accuracy: 0.7687\n", + "Epoch 00141: val_accuracy did not improve from 0.80206\n", + "7878/7878 [==============================] - 6s 800us/sample - loss: 0.5684 - accuracy: 0.7686 - val_loss: 0.6008 - val_accuracy: 0.7674\n", + "Epoch 142/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5537 - accuracy: 0.7736\n", + "Epoch 00142: val_accuracy improved from 0.80206 to 0.80891, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 810us/sample - loss: 0.5524 - accuracy: 0.7742 - val_loss: 0.4908 - val_accuracy: 0.8089\n", + "Epoch 143/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5560 - accuracy: 0.7735\n", + "Epoch 00143: val_accuracy did not improve from 0.80891\n", + "7878/7878 [==============================] - 6s 816us/sample - loss: 0.5565 - accuracy: 0.7733 - val_loss: 0.5076 - val_accuracy: 0.7937\n", + "Epoch 144/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5555 - accuracy: 0.7699\n", + "Epoch 00144: val_accuracy did not improve from 0.80891\n", + "7878/7878 [==============================] - 6s 808us/sample - loss: 0.5552 - accuracy: 0.7700 - val_loss: 0.5010 - val_accuracy: 0.7982\n", + "Epoch 145/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.5574 - accuracy: 0.7707\n", + "Epoch 00145: val_accuracy did not improve from 0.80891\n", + "7878/7878 [==============================] - 6s 811us/sample - loss: 0.5571 - accuracy: 0.7708 - val_loss: 0.5237 - val_accuracy: 0.7990\n", + "Epoch 146/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5559 - accuracy: 0.7744\n", + "Epoch 00146: val_accuracy did not improve from 0.80891\n", + "7878/7878 [==============================] - 7s 855us/sample - loss: 0.5558 - accuracy: 0.7744 - val_loss: 0.4987 - val_accuracy: 0.8047\n", + "Epoch 147/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5396 - accuracy: 0.7790\n", + "Epoch 00147: val_accuracy did not improve from 0.80891\n", + "7878/7878 [==============================] - 7s 833us/sample - loss: 0.5394 - accuracy: 0.7791 - val_loss: 0.4971 - val_accuracy: 0.8081\n", + "Epoch 148/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5552 - accuracy: 0.7701\n", + "Epoch 00148: val_accuracy did not improve from 0.80891\n", + "7878/7878 [==============================] - 7s 826us/sample - loss: 0.5554 - accuracy: 0.7699 - val_loss: 0.4884 - val_accuracy: 0.8062\n", + "Epoch 149/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5428 - accuracy: 0.7759\n", + "Epoch 00149: val_accuracy improved from 0.80891 to 0.80967, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 7s 825us/sample - loss: 0.5428 - accuracy: 0.7760 - val_loss: 0.4925 - val_accuracy: 0.8097\n", + "Epoch 150/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5427 - accuracy: 0.7798\n", + "Epoch 00150: val_accuracy did not improve from 0.80967\n", + "7878/7878 [==============================] - 6s 811us/sample - loss: 0.5428 - accuracy: 0.7800 - val_loss: 0.5069 - val_accuracy: 0.8051\n", + "Epoch 151/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.5390 - accuracy: 0.7846\n", + "Epoch 00151: val_accuracy did not improve from 0.80967\n", + "7878/7878 [==============================] - 6s 811us/sample - loss: 0.5397 - accuracy: 0.7838 - val_loss: 0.5047 - val_accuracy: 0.7910\n", + "Epoch 152/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5445 - accuracy: 0.7812\n", + "Epoch 00152: val_accuracy did not improve from 0.80967\n", + "7878/7878 [==============================] - 6s 816us/sample - loss: 0.5445 - accuracy: 0.7813 - val_loss: 0.5155 - val_accuracy: 0.7948\n", + "Epoch 153/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5435 - accuracy: 0.7763\n", + "Epoch 00153: val_accuracy improved from 0.80967 to 0.81271, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 820us/sample - loss: 0.5437 - accuracy: 0.7763 - val_loss: 0.5017 - val_accuracy: 0.8127\n", + "Epoch 154/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5302 - accuracy: 0.7806\n", + "Epoch 00154: val_accuracy did not improve from 0.81271\n", + "7878/7878 [==============================] - 6s 811us/sample - loss: 0.5299 - accuracy: 0.7808 - val_loss: 0.5754 - val_accuracy: 0.7845\n", + "Epoch 155/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5296 - accuracy: 0.7873\n", + "Epoch 00155: val_accuracy did not improve from 0.81271\n", + "7878/7878 [==============================] - 7s 907us/sample - loss: 0.5294 - accuracy: 0.7874 - val_loss: 0.4903 - val_accuracy: 0.8055\n", + "Epoch 156/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.5322 - accuracy: 0.7806\n", + "Epoch 00156: val_accuracy did not improve from 0.81271\n", + "7878/7878 [==============================] - 7s 890us/sample - loss: 0.5317 - accuracy: 0.7807 - val_loss: 0.4864 - val_accuracy: 0.7967\n", + "Epoch 157/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5306 - accuracy: 0.7814\n", + "Epoch 00157: val_accuracy did not improve from 0.81271\n", + "7878/7878 [==============================] - 6s 815us/sample - loss: 0.5312 - accuracy: 0.7809 - val_loss: 0.4886 - val_accuracy: 0.8059\n", + "Epoch 158/200\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.5294 - accuracy: 0.7816\n", + "Epoch 00158: val_accuracy did not improve from 0.81271\n", + "7878/7878 [==============================] - 6s 814us/sample - loss: 0.5295 - accuracy: 0.7817 - val_loss: 0.4905 - val_accuracy: 0.8017\n", + "Epoch 159/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.5216 - accuracy: 0.7824\n", + "Epoch 00159: val_accuracy did not improve from 0.81271\n", + "7878/7878 [==============================] - 6s 803us/sample - loss: 0.5209 - accuracy: 0.7827 - val_loss: 0.5032 - val_accuracy: 0.8028\n", + "Epoch 160/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.5287 - accuracy: 0.7837\n", + "Epoch 00160: val_accuracy did not improve from 0.81271\n", + "7878/7878 [==============================] - 6s 801us/sample - loss: 0.5285 - accuracy: 0.7836 - val_loss: 0.4860 - val_accuracy: 0.8120\n", + "Epoch 161/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5196 - accuracy: 0.7880\n", + "Epoch 00161: val_accuracy did not improve from 0.81271\n", + "7878/7878 [==============================] - 6s 807us/sample - loss: 0.5198 - accuracy: 0.7881 - val_loss: 0.4757 - val_accuracy: 0.8108\n", + "Epoch 162/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.5146 - accuracy: 0.7874\n", + "Epoch 00162: val_accuracy did not improve from 0.81271\n", + "7878/7878 [==============================] - 6s 794us/sample - loss: 0.5150 - accuracy: 0.7870 - val_loss: 0.4790 - val_accuracy: 0.8100\n", + "Epoch 163/200\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.5049 - accuracy: 0.7879\n", + "Epoch 00163: val_accuracy improved from 0.81271 to 0.81500, saving model to CNN_model_OLC.h5\n", + "7878/7878 [==============================] - 6s 808us/sample - loss: 0.5058 - accuracy: 0.7876 - val_loss: 0.4752 - val_accuracy: 0.8150\n", + "Epoch 164/200\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5139 - accuracy: 0.7879" + ] + } + ], + "source": [ + "history = model.fit(X_train, y_train, epochs=epoch, \n", + " validation_data=(X_cv, y_cv), \n", + " callbacks=callbacks,\n", + " verbose=1)" + ] + }, + { + "cell_type": "code", + "execution_count": 42, + "metadata": {}, + "outputs": [ + { + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "needs_background": "light" + }, + "output_type": "display_data" + } + ], + "source": [ + "def plot_acc_loss(history, epochs):\n", + " acc = history.history['accuracy']\n", + " loss = history.history['loss']\n", + " val_acc = history.history['val_accuracy']\n", + " val_loss = history.history['val_loss']\n", + " plt.figure(figsize=(20, 8))\n", + " plt.subplot(121)\n", + " plt.plot(range(1,epochs), acc[1:], label='Train Accuracy')\n", + " plt.plot(range(1,epochs), val_acc[1:], label='Test Accuracy')\n", + " plt.title('Total: ' + str(epochs) + ' runs; '+'ACCURACY', size=18)\n", + " plt.legend()\n", + " plt.grid(True)\n", + " plt.subplot(122)\n", + " plt.plot(range(1,epochs), loss[1:], label='Train Loss')\n", + " plt.plot(range(1,epochs), val_loss[1:], label='Test Loss')\n", + " plt.title('Total: ' + str(epochs) + ' runs; '+'LOSS', size=18)\n", + " plt.legend()\n", + " plt.grid(True)\n", + " plt.show()\n", + " \n", + "plot_acc_loss(history, 200)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "_cell_guid": "79c7e3d0-c299-4dcb-8224-4455121ee9b0", + "_uuid": "d629ff2d2480ee46fbb7e2d37f6b5fab8052498a" + }, + "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": "3.7.4" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +} diff --git a/dataAnalysis/LAHealthcareProject/DLmodels_CNN_OLC/CNNmulticlass_ZipCode.ipynb b/dataAnalysis/LAHealthcareProject/DLmodels_CNN_OLC/CNNmulticlass_ZipCode.ipynb new file mode 100644 index 000000000..a4bf9a208 --- /dev/null +++ b/dataAnalysis/LAHealthcareProject/DLmodels_CNN_OLC/CNNmulticlass_ZipCode.ipynb @@ -0,0 +1,1396 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Geospatial Analysis: Medically Underserved Areas based on Zip Code (Multi-Class CNN Model)\n", + "\n", + "**1. Data was first pre-processed and load the final version of la_facility_mua.csv here**\n", + "\n", + "**2. Since feature engineering has been done, a few touch-ups before training:** \n", + "\n", + "a) Feature Mapping,\n", + "\n", + "b) Dropping useless variables, \n", + "\n", + "c) Regex applications, \n", + "\n", + "d) Days Diff\n", + "\n", + "**3. Data was split into Train and Validation**\n", + "\n", + "**4. Model: Multi-Class CNN with Tensorflow.Keras (Zip Code: 51 classes)**\n", + "\n", + "**5. The best model with Train Loss: 0.1499 - Train Accuracy: 0.9335, Val Loss: 0.1008 - Val Accuracy: 0.9673**" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "os.environ['PYTHONHASHSEED']='0'" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "Using TensorFlow backend.\n" + ] + } + ], + "source": [ + "import random as rn\n", + "# Deep Learning Libraries\n", + "from tensorflow import keras\n", + "from tensorflow.keras import Sequential\n", + "from tensorflow.keras.layers import Flatten,Dense,Dropout,BatchNormalization\n", + "from tensorflow.keras.layers import Conv1D,MaxPool1D\n", + "from tensorflow.keras.optimizers import Adam\n", + "from tensorflow.keras.callbacks import EarlyStopping, ReduceLROnPlateau, ModelCheckpoint\n", + "from tensorflow.keras import callbacks\n", + "import tensorflow as tf\n", + "np.random.seed(37)\n", + "tf.random.set_seed(42)\n", + "from keras import backend as K" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "metadata": {}, + "outputs": [], + "source": [ + "#1\n", + "import pandas as pd\n", + "import numpy as np\n", + "import seaborn as sns\n", + "import re\n", + "from datetime import datetime\n", + "import matplotlib.pyplot as plt\n", + "%matplotlib inline\n", + "import matplotlib.gridspec as gridspec\n", + "from sklearn.preprocessing import StandardScaler\n", + "from sklearn.metrics import auc,roc_auc_score\n", + "from catboost import CatBoostClassifier,Pool, cv\n", + "from sklearn.preprocessing import LabelEncoder\n", + "from lightgbm import LGBMClassifier\n", + "from sklearn.metrics import roc_curve, auc, classification_report, confusion_matrix, accuracy_score\n", + "from sklearn.model_selection import train_test_split, cross_val_score, KFold, StratifiedKFold\n", + "\n", + "la_facility_mua = pd.read_csv('../input/la-final-data-2/la_facility_mua.csv')" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "metadata": {}, + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
sourcecat1cat2org_nameNameaddrln1citystatepost_iddescription...facility_address_1MUA_RuralStatusMUA_Designated_TimeMUA_Designated_DateMUA_MSSA_Name_1MUA_Updated_TimeMUA_Updated_DateMUA_PovertyMUA_ScoreMUA_Pct_100Poverty
0211Social ServicesChildren and Family Servicesneweconomicsforwomen.orgNew Economics For Women - Familysource Center ...501 S. Bixel St.Los AngelesCA56652The agency is an economic development organiza......415 South Union AvenueUrban1.067990e+122003-11-05MSSA 78.2b (Pico Union/Westlake)1.067990e+122003-11-05157757.051.00.0
1211Social ServicesChildren and Family Servicesneweconomicsforwomen.orgNew Economics For Women - Familysource Center ...501 S. Bixel St.Los AngelesCA56652The agency is an economic development organiza......600 St Paul AveUrban7.686144e+111994-05-11Westlake Service Area1.005782e+122001-11-150.060.10.0
2211Social ServicesChildren and Family Servicesneweconomicsforwomen.orgNew Economics For Women - Familysource Center ...501 S. Bixel St.Los AngelesCA56652The agency is an economic development organiza......1125 W 6th StUrban7.686144e+111994-05-11Westlake Service Area1.005782e+122001-11-150.060.10.0
3211Social ServicesChildren and Family Servicesneweconomicsforwomen.orgNew Economics For Women - Familysource Center ...501 S. Bixel St.Los AngelesCA56652The agency is an economic development organiza......1225 Wilshire BoulevardUrban7.686144e+111994-05-11Westlake Service Area1.005782e+122001-11-150.060.10.0
4211Social ServicesChildren and Family Servicesneweconomicsforwomen.orgNew Economics For Women - Familysource Center ...501 S. Bixel St.Los AngelesCA56652The agency is an economic development organiza......1127 Wilshire BlvdUrban7.686144e+111994-05-11Westlake Service Area1.005782e+122001-11-150.060.10.0
5211Social ServicesChildren and Family Servicesneweconomicsforwomen.orgNew Economics For Women - Familysource Center ...501 S. Bixel St.Los AngelesCA56652The agency is an economic development organiza......1212 Wilshire BlvdUrban7.686144e+111994-05-11Westlake Service Area1.005782e+122001-11-150.060.10.0
\n", + "

6 rows × 32 columns

\n", + "
" + ], + "text/plain": [ + " source cat1 cat2 \\\n", + "0 211 Social Services Children and Family Services \n", + "1 211 Social Services Children and Family Services \n", + "2 211 Social Services Children and Family Services \n", + "3 211 Social Services Children and Family Services \n", + "4 211 Social Services Children and Family Services \n", + "5 211 Social Services Children and Family Services \n", + "\n", + " org_name \\\n", + "0 neweconomicsforwomen.org \n", + "1 neweconomicsforwomen.org \n", + "2 neweconomicsforwomen.org \n", + "3 neweconomicsforwomen.org \n", + "4 neweconomicsforwomen.org \n", + "5 neweconomicsforwomen.org \n", + "\n", + " Name addrln1 \\\n", + "0 New Economics For Women - Familysource Center ... 501 S. Bixel St. \n", + "1 New Economics For Women - Familysource Center ... 501 S. Bixel St. \n", + "2 New Economics For Women - Familysource Center ... 501 S. Bixel St. \n", + "3 New Economics For Women - Familysource Center ... 501 S. Bixel St. \n", + "4 New Economics For Women - Familysource Center ... 501 S. Bixel St. \n", + "5 New Economics For Women - Familysource Center ... 501 S. Bixel St. \n", + "\n", + " city state post_id \\\n", + "0 Los Angeles CA 56652 \n", + "1 Los Angeles CA 56652 \n", + "2 Los Angeles CA 56652 \n", + "3 Los Angeles CA 56652 \n", + "4 Los Angeles CA 56652 \n", + "5 Los Angeles CA 56652 \n", + "\n", + " description ... \\\n", + "0 The agency is an economic development organiza... ... \n", + "1 The agency is an economic development organiza... ... \n", + "2 The agency is an economic development organiza... ... \n", + "3 The agency is an economic development organiza... ... \n", + "4 The agency is an economic development organiza... ... \n", + "5 The agency is an economic development organiza... ... \n", + "\n", + " facility_address_1 MUA_RuralStatus MUA_Designated_Time \\\n", + "0 415 South Union Avenue Urban 1.067990e+12 \n", + "1 600 St Paul Ave Urban 7.686144e+11 \n", + "2 1125 W 6th St Urban 7.686144e+11 \n", + "3 1225 Wilshire Boulevard Urban 7.686144e+11 \n", + "4 1127 Wilshire Blvd Urban 7.686144e+11 \n", + "5 1212 Wilshire Blvd Urban 7.686144e+11 \n", + "\n", + " MUA_Designated_Date MUA_MSSA_Name_1 MUA_Updated_Time \\\n", + "0 2003-11-05 MSSA 78.2b (Pico Union/Westlake) 1.067990e+12 \n", + "1 1994-05-11 Westlake Service Area 1.005782e+12 \n", + "2 1994-05-11 Westlake Service Area 1.005782e+12 \n", + "3 1994-05-11 Westlake Service Area 1.005782e+12 \n", + "4 1994-05-11 Westlake Service Area 1.005782e+12 \n", + "5 1994-05-11 Westlake Service Area 1.005782e+12 \n", + "\n", + " MUA_Updated_Date MUA_Poverty MUA_Score MUA_Pct_100Poverty \n", + "0 2003-11-05 157757.0 51.0 0.0 \n", + "1 2001-11-15 0.0 60.1 0.0 \n", + "2 2001-11-15 0.0 60.1 0.0 \n", + "3 2001-11-15 0.0 60.1 0.0 \n", + "4 2001-11-15 0.0 60.1 0.0 \n", + "5 2001-11-15 0.0 60.1 0.0 \n", + "\n", + "[6 rows x 32 columns]" + ] + }, + "execution_count": 5, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "#2\n", + "la_facility_mua.head(6)" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "metadata": {}, + "outputs": [], + "source": [ + "#3\n", + "la_facility_mua=la_facility_mua[la_facility_mua['MUA_Designated_Time'].notnull()]\n", + "la_facility_mua.drop(['facility_number_of_beds', 'facility_name', 'org_name', 'Name'],\n", + " axis=1, inplace=True)\n", + "la_facility_mua= la_facility_mua[la_facility_mua['Zip Code']!=0.0]\n", + "la_facility_mua['Zip Code']=la_facility_mua['Zip Code'].astype('int64')" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "metadata": {}, + "outputs": [], + "source": [ + "#4\n", + "la_facility_mua['facility_er_service'] = la_facility_mua['facility_er_service'].map({\n", + "'Not Applicable':'Not Applicable',\n", + "'None':'Not Applicable',\n", + "'Emergency - Basic':'Emergency',\n", + "'Emergency - Comprehensive':'Emergency',\n", + "'Emergency - Standby':'Emergency'\n", + "})" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "metadata": {}, + "outputs": [], + "source": [ + "#5\n", + "la_facility_mua['MUA_Poverty']=la_facility_mua['MUA_Poverty'].map({\n", + "157757.0:2,\n", + "129447.0:1,\n", + "0.0:0,\n", + "})" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "metadata": {}, + "outputs": [], + "source": [ + "#6\n", + "la_facility_mua['cat1+2'] = la_facility_mua['cat2'].astype('str')+'_'+la_facility_mua['cat1'].astype('str')\n", + "\n", + "la_facility_mua['city'] = la_facility_mua['city'].str.replace('Signal Hill ', 'Signal Hill')\n", + "la_facility_mua['city'] = la_facility_mua['city'].str.replace(' Hollywoo', ' Hollywood').str.replace(' Hollywoodd', ' Hollywood')\n", + "la_facility_mua['city'] = la_facility_mua['city'].str.replace('Los Angeles,', 'Los Angeles').str.replace('Los Angeles ', 'Los Angeles').str.replace('Los Angelels','Los Angeles').str.replace('Los Angeles', 'Los Angeles').str.replace(' Los Angeles', 'Los Angeles')\n", + "la_facility_mua['city'] = la_facility_mua['city'].str.replace('Industry', 'City of Industry').str.replace('City of Industry ', 'City of Industry').str.replace('City of City of Industry', 'City of Industry')\n", + "la_facility_mua['city'] = la_facility_mua['city'].str.replace('Bell', 'Bell Gardens').str.replace('Bell Gardensflower', 'Bell Gardens').str.replace('Bell Gardens Gardens', 'Bell Gardens')" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "metadata": {}, + "outputs": [], + "source": [ + "#7\n", + "la_facility_mua['Address'] = la_facility_mua['facility_address_1'].str.lower()\n", + "la_facility_mua['Street'] = la_facility_mua['Address'].str.extract('([a-zA-Z ]+)', expand=False).str.strip()\n", + "\n", + "la_facility_mua['Street'] = la_facility_mua['Street'].str.replace(' boulevard', ' blvd').str.replace(' place',' pl').str.replace(' avenue',' ave').str.replace(' street',' st').str.replace(' highway',' hwy').str.replace(' road',' rd').str.replace(' drive',' dr').str.replace(' huntington drive south', ' south huntington dr').str.replace(' huntington dr n', ' north huntington dr').str.replace('south ', 's ').str.replace('north ','n ').str.replace('east ','e ').str.replace('west ','w ').str.replace('south','s').str.replace('north','n').str.replace('east','e').str.replace('west','w')" + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "metadata": {}, + "outputs": [], + "source": [ + "#8\n", + "la_facility_mua['MUA_Designated_Date'] = pd.to_datetime(la_facility_mua['MUA_Designated_Date']) \n", + "la_facility_mua['MUA_Updated_Date'] = pd.to_datetime(la_facility_mua['MUA_Updated_Date']) \n", + "la_facility_mua['Date Difference'] = la_facility_mua['MUA_Updated_Date']-la_facility_mua['MUA_Designated_Date']" + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "metadata": {}, + "outputs": [], + "source": [ + "#9\n", + "def ddiff2days(ddiff):\n", + " if not pd.isnull(ddiff):\n", + " return pd.Timedelta.total_seconds(ddiff)/(24.*3600)\n", + " else:\n", + " return np.NaN" + ] + }, + { + "cell_type": "code", + "execution_count": 18, + "metadata": {}, + "outputs": [], + "source": [ + "#10\n", + "la_facility_mua['Date Difference']=la_facility_mua['Date Difference'].apply(ddiff2days)\n", + "la_facility_mua['Date Difference']=la_facility_mua['Date Difference'].astype('int32')" + ] + }, + { + "cell_type": "code", + "execution_count": 19, + "metadata": {}, + "outputs": [], + "source": [ + "#11\n", + "la_facility_mua['design_year'] = (la_facility_mua['MUA_Designated_Date'].dt.year).astype('int32')\n", + "la_facility_mua['design_month'] = (la_facility_mua['MUA_Designated_Date'].dt.month).astype('int32')\n", + "la_facility_mua['design_week'] = (la_facility_mua['MUA_Designated_Date'].dt.week).astype('int32')\n", + "la_facility_mua['design_dow'] = (la_facility_mua['MUA_Designated_Date'].dt.dayofweek).astype('int32')\n", + "\n", + "la_facility_mua['update_year'] = (la_facility_mua['MUA_Updated_Date'].dt.year).astype('int32')\n", + "la_facility_mua['update_month'] = (la_facility_mua['MUA_Updated_Date'].dt.month).astype('int32')\n", + "la_facility_mua['update_week'] = (la_facility_mua['MUA_Updated_Date'].dt.week).astype('int32')\n", + "la_facility_mua['update_dow'] = (la_facility_mua['MUA_Updated_Date'].dt.dayofweek).astype('int32')" + ] + }, + { + "cell_type": "code", + "execution_count": 22, + "metadata": {}, + "outputs": [], + "source": [ + "#12\n", + "la_facility_mua['Zip Code']=la_facility_mua['Zip Code'].astype('object')" + ] + }, + { + "cell_type": "code", + "execution_count": 31, + "metadata": {}, + "outputs": [], + "source": [ + "#13\n", + "la_facility_mua.drop(['MUA_Designated_Date', 'MUA_Updated_Date', 'facility_address_1'\n", + " ], axis=1, inplace=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 32, + "metadata": {}, + "outputs": [], + "source": [ + "#14\n", + "cols = [c for c in la_facility_mua.columns if la_facility_mua[c].dtypes=='object']\n", + "for c in cols:\n", + " le = LabelEncoder()\n", + " la_facility_mua[c] = (le.fit_transform(la_facility_mua[c])).astype('int32')" + ] + }, + { + "cell_type": "code", + "execution_count": 33, + "metadata": {}, + "outputs": [], + "source": [ + "#15\n", + "X = la_facility_mua.drop(labels=['Zip Code'], axis=1)\n", + "y = la_facility_mua['Zip Code'].values\n", + "\n", + "from sklearn.model_selection import train_test_split\n", + "X_train, X_cv, y_train, y_cv = train_test_split(X, y, test_size=0.25, random_state=tf.random.set_seed(42))" + ] + }, + { + "cell_type": "code", + "execution_count": 34, + "metadata": {}, + "outputs": [], + "source": [ + "#16\n", + "X_train = X_train.to_numpy()\n", + "X_cv = X_cv.to_numpy()" + ] + }, + { + "cell_type": "code", + "execution_count": 35, + "metadata": {}, + "outputs": [], + "source": [ + "#17\n", + "X_train = X_train.reshape(X_train.shape[0],X_train.shape[1],1)\n", + "X_cv = X_cv.reshape(X_cv.shape[0],X_cv.shape[1],1)" + ] + }, + { + "cell_type": "code", + "execution_count": 36, + "metadata": {}, + "outputs": [], + "source": [ + "#18\n", + "# Convolutional Neural Networks (CNN)\n", + "epoch = 180\n", + "model = Sequential()\n", + "model.add(Conv1D(40,2,activation='relu',kernel_initializer='he_normal',input_shape = X_train[0].shape))\n", + "model.add(Conv1D(100,2,activation='relu', kernel_initializer='he_normal'))\n", + "model.add(Conv1D(70,2,activation='relu', kernel_initializer='he_normal'))\n", + "model.add(BatchNormalization())\n", + "model.add(Dropout(0.4))\n", + "\n", + "model.add(Flatten())\n", + "model.add(Dense(50,activation='relu', kernel_initializer='he_normal'))\n", + "model.add(Dense(51,activation='softmax'))" + ] + }, + { + "cell_type": "code", + "execution_count": 37, + "metadata": {}, + "outputs": [], + "source": [ + "#19\n", + "model.compile(optimizer=Adam(lr=0.001), loss = 'sparse_categorical_crossentropy', \n", + " metrics=['accuracy'])" + ] + }, + { + "cell_type": "code", + "execution_count": 38, + "metadata": {}, + "outputs": [], + "source": [ + "#20\n", + "earlystop = EarlyStopping(patience=20) \n", + "\n", + "learning_rate_reduction = ReduceLROnPlateau(monitor='val_accuracy', \n", + " patience=5, \n", + " verbose=1, \n", + " factor=0.96, \n", + " min_lr=0.00001)\n", + "checkpointer = ModelCheckpoint(filepath='CNN_model_2.h5', monitor='val_accuracy', verbose=1, \n", + " save_best_only=True, mode='max', period=1)\n", + "callbacks = [#earlystop, \n", + " learning_rate_reduction, checkpointer\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": 39, + "metadata": { + "scrolled": false + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Train on 7878 samples, validate on 2627 samples\n", + "Epoch 1/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 3.2782 - accuracy: 0.1874\n", + "Epoch 00001: val_accuracy improved from -inf to 0.30872, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 7s 833us/sample - loss: 3.2727 - accuracy: 0.1880 - val_loss: 2.5220 - val_accuracy: 0.3087\n", + "Epoch 2/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 2.1745 - accuracy: 0.3609\n", + "Epoch 00002: val_accuracy improved from 0.30872 to 0.45641, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 606us/sample - loss: 2.1710 - accuracy: 0.3614 - val_loss: 1.6232 - val_accuracy: 0.4564\n", + "Epoch 3/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.6884 - accuracy: 0.4431\n", + "Epoch 00003: val_accuracy improved from 0.45641 to 0.50476, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 613us/sample - loss: 1.6873 - accuracy: 0.4433 - val_loss: 1.3993 - val_accuracy: 0.5048\n", + "Epoch 4/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.4762 - accuracy: 0.4769\n", + "Epoch 00004: val_accuracy improved from 0.50476 to 0.53940, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 605us/sample - loss: 1.4756 - accuracy: 0.4770 - val_loss: 1.2102 - val_accuracy: 0.5394\n", + "Epoch 5/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.3237 - accuracy: 0.5190\n", + "Epoch 00005: val_accuracy improved from 0.53940 to 0.59079, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 604us/sample - loss: 1.3241 - accuracy: 0.5182 - val_loss: 1.0638 - val_accuracy: 0.5908\n", + "Epoch 6/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 1.2305 - accuracy: 0.5418\n", + "Epoch 00006: val_accuracy improved from 0.59079 to 0.61096, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 637us/sample - loss: 1.2304 - accuracy: 0.5418 - val_loss: 1.0327 - val_accuracy: 0.6110\n", + "Epoch 7/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.1465 - accuracy: 0.5678\n", + "Epoch 00007: val_accuracy improved from 0.61096 to 0.64332, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 615us/sample - loss: 1.1454 - accuracy: 0.5679 - val_loss: 0.9498 - val_accuracy: 0.6433\n", + "Epoch 8/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.0773 - accuracy: 0.5936\n", + "Epoch 00008: val_accuracy improved from 0.64332 to 0.64484, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 655us/sample - loss: 1.0786 - accuracy: 0.5927 - val_loss: 0.8870 - val_accuracy: 0.6448\n", + "Epoch 9/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 1.0313 - accuracy: 0.6039\n", + "Epoch 00009: val_accuracy improved from 0.64484 to 0.65207, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 657us/sample - loss: 1.0310 - accuracy: 0.6038 - val_loss: 0.8982 - val_accuracy: 0.6521\n", + "Epoch 10/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.9835 - accuracy: 0.6296\n", + "Epoch 00010: val_accuracy improved from 0.65207 to 0.70917, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 611us/sample - loss: 0.9827 - accuracy: 0.6295 - val_loss: 0.8166 - val_accuracy: 0.7092\n", + "Epoch 11/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.9351 - accuracy: 0.6372\n", + "Epoch 00011: val_accuracy did not improve from 0.70917\n", + "7878/7878 [==============================] - 5s 600us/sample - loss: 0.9355 - accuracy: 0.6372 - val_loss: 0.7733 - val_accuracy: 0.6981\n", + "Epoch 12/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.9099 - accuracy: 0.6425\n", + "Epoch 00012: val_accuracy did not improve from 0.70917\n", + "7878/7878 [==============================] - 5s 595us/sample - loss: 0.9112 - accuracy: 0.6420 - val_loss: 0.7713 - val_accuracy: 0.6970\n", + "Epoch 13/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.8923 - accuracy: 0.6481\n", + "Epoch 00013: val_accuracy did not improve from 0.70917\n", + "7878/7878 [==============================] - 5s 599us/sample - loss: 0.8922 - accuracy: 0.6478 - val_loss: 0.7228 - val_accuracy: 0.7004\n", + "Epoch 14/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.8633 - accuracy: 0.6577\n", + "Epoch 00014: val_accuracy improved from 0.70917 to 0.71983, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 595us/sample - loss: 0.8651 - accuracy: 0.6570 - val_loss: 0.7030 - val_accuracy: 0.7198\n", + "Epoch 15/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.8447 - accuracy: 0.6644\n", + "Epoch 00015: val_accuracy did not improve from 0.71983\n", + "7878/7878 [==============================] - 5s 603us/sample - loss: 0.8462 - accuracy: 0.6637 - val_loss: 0.7138 - val_accuracy: 0.7008\n", + "Epoch 16/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.8264 - accuracy: 0.6801\n", + "Epoch 00016: val_accuracy improved from 0.71983 to 0.73925, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 610us/sample - loss: 0.8264 - accuracy: 0.6794 - val_loss: 0.6706 - val_accuracy: 0.7392\n", + "Epoch 17/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.8008 - accuracy: 0.6847\n", + "Epoch 00017: val_accuracy did not improve from 0.73925\n", + "7878/7878 [==============================] - 5s 595us/sample - loss: 0.8013 - accuracy: 0.6849 - val_loss: 0.6495 - val_accuracy: 0.7381\n", + "Epoch 18/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.7787 - accuracy: 0.6916\n", + "Epoch 00018: val_accuracy improved from 0.73925 to 0.75866, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 604us/sample - loss: 0.7788 - accuracy: 0.6913 - val_loss: 0.6104 - val_accuracy: 0.7587\n", + "Epoch 19/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.7697 - accuracy: 0.6948\n", + "Epoch 00019: val_accuracy improved from 0.75866 to 0.77389, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 601us/sample - loss: 0.7688 - accuracy: 0.6951 - val_loss: 0.5907 - val_accuracy: 0.7739\n", + "Epoch 20/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.7416 - accuracy: 0.7092\n", + "Epoch 00020: val_accuracy did not improve from 0.77389\n", + "7878/7878 [==============================] - 5s 632us/sample - loss: 0.7417 - accuracy: 0.7092 - val_loss: 0.5892 - val_accuracy: 0.7632\n", + "Epoch 21/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.7265 - accuracy: 0.7138\n", + "Epoch 00021: val_accuracy improved from 0.77389 to 0.77617, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 637us/sample - loss: 0.7264 - accuracy: 0.7139 - val_loss: 0.5579 - val_accuracy: 0.7762\n", + "Epoch 22/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.7012 - accuracy: 0.7205\n", + "Epoch 00022: val_accuracy improved from 0.77617 to 0.77998, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 613us/sample - loss: 0.7010 - accuracy: 0.7207 - val_loss: 0.5532 - val_accuracy: 0.7800\n", + "Epoch 23/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6992 - accuracy: 0.7205\n", + "Epoch 00023: val_accuracy improved from 0.77998 to 0.79292, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 611us/sample - loss: 0.6988 - accuracy: 0.7211 - val_loss: 0.5341 - val_accuracy: 0.7929\n", + "Epoch 24/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6641 - accuracy: 0.7335\n", + "Epoch 00024: val_accuracy improved from 0.79292 to 0.81386, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 604us/sample - loss: 0.6637 - accuracy: 0.7337 - val_loss: 0.4882 - val_accuracy: 0.8139\n", + "Epoch 25/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6603 - accuracy: 0.7314\n", + "Epoch 00025: val_accuracy improved from 0.81386 to 0.82185, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 606us/sample - loss: 0.6597 - accuracy: 0.7322 - val_loss: 0.4819 - val_accuracy: 0.8219\n", + "Epoch 26/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6283 - accuracy: 0.7532\n", + "Epoch 00026: val_accuracy did not improve from 0.82185\n", + "7878/7878 [==============================] - 5s 595us/sample - loss: 0.6287 - accuracy: 0.7525 - val_loss: 0.4995 - val_accuracy: 0.8127\n", + "Epoch 27/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.6137 - accuracy: 0.7591\n", + "Epoch 00027: val_accuracy did not improve from 0.82185\n", + "7878/7878 [==============================] - 5s 600us/sample - loss: 0.6138 - accuracy: 0.7593 - val_loss: 0.5130 - val_accuracy: 0.8002\n", + "Epoch 28/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5797 - accuracy: 0.7704\n", + "Epoch 00028: val_accuracy improved from 0.82185 to 0.82223, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 605us/sample - loss: 0.5793 - accuracy: 0.7702 - val_loss: 0.4387 - val_accuracy: 0.8222\n", + "Epoch 29/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5749 - accuracy: 0.7741\n", + "Epoch 00029: val_accuracy improved from 0.82223 to 0.82908, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 608us/sample - loss: 0.5748 - accuracy: 0.7737 - val_loss: 0.4322 - val_accuracy: 0.8291\n", + "Epoch 30/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5598 - accuracy: 0.7769\n", + "Epoch 00030: val_accuracy did not improve from 0.82908\n", + "7878/7878 [==============================] - 5s 608us/sample - loss: 0.5599 - accuracy: 0.7771 - val_loss: 0.4481 - val_accuracy: 0.8207\n", + "Epoch 31/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5447 - accuracy: 0.7830\n", + "Epoch 00031: val_accuracy improved from 0.82908 to 0.83099, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 608us/sample - loss: 0.5453 - accuracy: 0.7829 - val_loss: 0.4289 - val_accuracy: 0.8310\n", + "Epoch 32/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5382 - accuracy: 0.7889\n", + "Epoch 00032: val_accuracy improved from 0.83099 to 0.84888, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 620us/sample - loss: 0.5367 - accuracy: 0.7892 - val_loss: 0.3799 - val_accuracy: 0.8489\n", + "Epoch 33/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5374 - accuracy: 0.7843\n", + "Epoch 00033: val_accuracy did not improve from 0.84888\n", + "7878/7878 [==============================] - 5s 633us/sample - loss: 0.5382 - accuracy: 0.7838 - val_loss: 0.3600 - val_accuracy: 0.8382\n", + "Epoch 34/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5100 - accuracy: 0.8043\n", + "Epoch 00034: val_accuracy did not improve from 0.84888\n", + "7878/7878 [==============================] - 5s 613us/sample - loss: 0.5094 - accuracy: 0.8045 - val_loss: 0.3693 - val_accuracy: 0.8489\n", + "Epoch 35/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.5021 - accuracy: 0.8007\n", + "Epoch 00035: val_accuracy improved from 0.84888 to 0.85687, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 605us/sample - loss: 0.5023 - accuracy: 0.8006 - val_loss: 0.3691 - val_accuracy: 0.8569\n", + "Epoch 36/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.4996 - accuracy: 0.7976\n", + "Epoch 00036: val_accuracy did not improve from 0.85687\n", + "7878/7878 [==============================] - 5s 611us/sample - loss: 0.4997 - accuracy: 0.7974 - val_loss: 0.3593 - val_accuracy: 0.8515\n", + "Epoch 37/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.4721 - accuracy: 0.8117\n", + "Epoch 00037: val_accuracy improved from 0.85687 to 0.86220, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 608us/sample - loss: 0.4739 - accuracy: 0.8109 - val_loss: 0.3542 - val_accuracy: 0.8622\n", + "Epoch 38/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.4659 - accuracy: 0.8139\n", + "Epoch 00038: val_accuracy improved from 0.86220 to 0.87057, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 615us/sample - loss: 0.4657 - accuracy: 0.8139 - val_loss: 0.3247 - val_accuracy: 0.8706\n", + "Epoch 39/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.4624 - accuracy: 0.8183\n", + "Epoch 00039: val_accuracy improved from 0.87057 to 0.87895, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 605us/sample - loss: 0.4634 - accuracy: 0.8180 - val_loss: 0.3232 - val_accuracy: 0.8789\n", + "Epoch 40/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.4535 - accuracy: 0.8202\n", + "Epoch 00040: val_accuracy did not improve from 0.87895\n", + "7878/7878 [==============================] - 5s 594us/sample - loss: 0.4536 - accuracy: 0.8201 - val_loss: 0.3189 - val_accuracy: 0.8698\n", + "Epoch 41/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.4570 - accuracy: 0.8203\n", + "Epoch 00041: val_accuracy did not improve from 0.87895\n", + "7878/7878 [==============================] - 5s 601us/sample - loss: 0.4573 - accuracy: 0.8203 - val_loss: 0.3145 - val_accuracy: 0.8660\n", + "Epoch 42/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.4277 - accuracy: 0.8308\n", + "Epoch 00042: val_accuracy did not improve from 0.87895\n", + "7878/7878 [==============================] - 5s 594us/sample - loss: 0.4279 - accuracy: 0.8304 - val_loss: 0.3131 - val_accuracy: 0.8736\n", + "Epoch 43/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.4484 - accuracy: 0.8215\n", + "Epoch 00043: val_accuracy improved from 0.87895 to 0.88847, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 612us/sample - loss: 0.4477 - accuracy: 0.8218 - val_loss: 0.3027 - val_accuracy: 0.8885\n", + "Epoch 44/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.4236 - accuracy: 0.8344\n", + "Epoch 00044: val_accuracy did not improve from 0.88847\n", + "7878/7878 [==============================] - 5s 593us/sample - loss: 0.4230 - accuracy: 0.8351 - val_loss: 0.3035 - val_accuracy: 0.8770\n", + "Epoch 45/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.4195 - accuracy: 0.8290\n", + "Epoch 00045: val_accuracy did not improve from 0.88847\n", + "7878/7878 [==============================] - 5s 637us/sample - loss: 0.4194 - accuracy: 0.8290 - val_loss: 0.3638 - val_accuracy: 0.8557\n", + "Epoch 46/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.4221 - accuracy: 0.8348\n", + "Epoch 00046: val_accuracy improved from 0.88847 to 0.90369, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 629us/sample - loss: 0.4223 - accuracy: 0.8350 - val_loss: 0.2750 - val_accuracy: 0.9037\n", + "Epoch 47/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.4050 - accuracy: 0.8368\n", + "Epoch 00047: val_accuracy did not improve from 0.90369\n", + "7878/7878 [==============================] - 5s 653us/sample - loss: 0.4052 - accuracy: 0.8368 - val_loss: 0.2716 - val_accuracy: 0.8976\n", + "Epoch 48/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.4000 - accuracy: 0.8412\n", + "Epoch 00048: val_accuracy did not improve from 0.90369\n", + "7878/7878 [==============================] - 5s 658us/sample - loss: 0.4004 - accuracy: 0.8409 - val_loss: 0.2999 - val_accuracy: 0.8789\n", + "Epoch 49/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.3954 - accuracy: 0.8401\n", + "Epoch 00049: val_accuracy improved from 0.90369 to 0.90445, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 616us/sample - loss: 0.3971 - accuracy: 0.8397 - val_loss: 0.2499 - val_accuracy: 0.9045\n", + "Epoch 50/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.3948 - accuracy: 0.8431\n", + "Epoch 00050: val_accuracy did not improve from 0.90445\n", + "7878/7878 [==============================] - 5s 609us/sample - loss: 0.3950 - accuracy: 0.8430 - val_loss: 0.2896 - val_accuracy: 0.8949\n", + "Epoch 51/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.3917 - accuracy: 0.8438\n", + "Epoch 00051: val_accuracy improved from 0.90445 to 0.91435, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 598us/sample - loss: 0.3905 - accuracy: 0.8442 - val_loss: 0.2446 - val_accuracy: 0.9144\n", + "Epoch 52/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.3735 - accuracy: 0.8537\n", + "Epoch 00052: val_accuracy did not improve from 0.91435\n", + "7878/7878 [==============================] - 5s 595us/sample - loss: 0.3736 - accuracy: 0.8538 - val_loss: 0.2418 - val_accuracy: 0.9010\n", + "Epoch 53/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.3774 - accuracy: 0.8489\n", + "Epoch 00053: val_accuracy did not improve from 0.91435\n", + "7878/7878 [==============================] - 5s 593us/sample - loss: 0.3767 - accuracy: 0.8489 - val_loss: 0.3287 - val_accuracy: 0.8793\n", + "Epoch 54/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.3687 - accuracy: 0.8538\n", + "Epoch 00054: val_accuracy did not improve from 0.91435\n", + "7878/7878 [==============================] - 5s 594us/sample - loss: 0.3689 - accuracy: 0.8538 - val_loss: 0.2790 - val_accuracy: 0.8999\n", + "Epoch 55/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.3652 - accuracy: 0.8587\n", + "Epoch 00055: val_accuracy did not improve from 0.91435\n", + "7878/7878 [==============================] - 5s 596us/sample - loss: 0.3651 - accuracy: 0.8586 - val_loss: 0.2553 - val_accuracy: 0.9022\n", + "Epoch 56/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.3553 - accuracy: 0.8589\n", + "Epoch 00056: ReduceLROnPlateau reducing learning rate to 0.0009600000455975532.\n", + "\n", + "Epoch 00056: val_accuracy did not improve from 0.91435\n", + "7878/7878 [==============================] - 5s 597us/sample - loss: 0.3557 - accuracy: 0.8587 - val_loss: 0.2503 - val_accuracy: 0.9014\n", + "Epoch 57/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.3661 - accuracy: 0.8485\n", + "Epoch 00057: val_accuracy did not improve from 0.91435\n", + "7878/7878 [==============================] - 5s 632us/sample - loss: 0.3667 - accuracy: 0.8483 - val_loss: 0.2342 - val_accuracy: 0.9048\n", + "Epoch 58/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.3444 - accuracy: 0.8592\n", + "Epoch 00058: val_accuracy did not improve from 0.91435\n", + "7878/7878 [==============================] - 5s 625us/sample - loss: 0.3447 - accuracy: 0.8587 - val_loss: 0.2359 - val_accuracy: 0.9136\n", + "Epoch 59/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2924 - accuracy: 0.8843\n", + "Epoch 00079: val_accuracy did not improve from 0.93148\n", + "7878/7878 [==============================] - 5s 591us/sample - loss: 0.2923 - accuracy: 0.8845 - val_loss: 0.2017 - val_accuracy: 0.9242\n", + "Epoch 80/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2822 - accuracy: 0.8863\n", + "Epoch 00080: val_accuracy did not improve from 0.93148\n", + "7878/7878 [==============================] - 5s 596us/sample - loss: 0.2826 - accuracy: 0.8860 - val_loss: 0.1928 - val_accuracy: 0.9242\n", + "Epoch 81/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2864 - accuracy: 0.8847\n", + "Epoch 00081: val_accuracy did not improve from 0.93148\n", + "7878/7878 [==============================] - 5s 593us/sample - loss: 0.2863 - accuracy: 0.8845 - val_loss: 0.2085 - val_accuracy: 0.9292\n", + "Epoch 82/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2770 - accuracy: 0.8886\n", + "Epoch 00082: ReduceLROnPlateau reducing learning rate to 0.000884736031293869.\n", + "\n", + "Epoch 00082: val_accuracy did not improve from 0.93148\n", + "7878/7878 [==============================] - 5s 598us/sample - loss: 0.2761 - accuracy: 0.8891 - val_loss: 0.1912 - val_accuracy: 0.9201\n", + "Epoch 83/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.2767 - accuracy: 0.8848\n", + "Epoch 00083: val_accuracy did not improve from 0.93148\n", + "7878/7878 [==============================] - 5s 636us/sample - loss: 0.2775 - accuracy: 0.8845 - val_loss: 0.1909 - val_accuracy: 0.9220\n", + "Epoch 84/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.2811 - accuracy: 0.8849\n", + "Epoch 00084: val_accuracy improved from 0.93148 to 0.93186, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 644us/sample - loss: 0.2815 - accuracy: 0.8847 - val_loss: 0.1920 - val_accuracy: 0.9319\n", + "Epoch 85/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2700 - accuracy: 0.8909\n", + "Epoch 00085: val_accuracy did not improve from 0.93186\n", + "7878/7878 [==============================] - 5s 594us/sample - loss: 0.2698 - accuracy: 0.8913 - val_loss: 0.1865 - val_accuracy: 0.9296\n", + "Epoch 86/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2786 - accuracy: 0.8900\n", + "Epoch 00086: val_accuracy did not improve from 0.93186\n", + "7878/7878 [==============================] - 5s 595us/sample - loss: 0.2788 - accuracy: 0.8901 - val_loss: 0.1858 - val_accuracy: 0.9235\n", + "Epoch 87/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.2812 - accuracy: 0.8908\n", + "Epoch 00087: val_accuracy improved from 0.93186 to 0.94062, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 601us/sample - loss: 0.2810 - accuracy: 0.8908 - val_loss: 0.1683 - val_accuracy: 0.9406\n", + "Epoch 88/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.2710 - accuracy: 0.8918\n", + "Epoch 00088: val_accuracy did not improve from 0.94062\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.2715 - accuracy: 0.8920 - val_loss: 0.1681 - val_accuracy: 0.9281\n", + "Epoch 89/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.2771 - accuracy: 0.8900\n", + "Epoch 00089: val_accuracy did not improve from 0.94062\n", + "7878/7878 [==============================] - 5s 597us/sample - loss: 0.2770 - accuracy: 0.8901 - val_loss: 0.1647 - val_accuracy: 0.9368\n", + "Epoch 90/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2731 - accuracy: 0.8893\n", + "Epoch 00090: val_accuracy did not improve from 0.94062\n", + "7878/7878 [==============================] - 5s 593us/sample - loss: 0.2721 - accuracy: 0.8897 - val_loss: 0.1858 - val_accuracy: 0.9277\n", + "Epoch 91/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.2544 - accuracy: 0.8955\n", + "Epoch 00091: val_accuracy did not improve from 0.94062\n", + "7878/7878 [==============================] - 5s 607us/sample - loss: 0.2546 - accuracy: 0.8955 - val_loss: 0.1834 - val_accuracy: 0.9322\n", + "Epoch 92/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2538 - accuracy: 0.8974\n", + "Epoch 00092: ReduceLROnPlateau reducing learning rate to 0.000849346574395895.\n", + "\n", + "Epoch 00092: val_accuracy did not improve from 0.94062\n", + "7878/7878 [==============================] - 5s 591us/sample - loss: 0.2545 - accuracy: 0.8971 - val_loss: 0.1777 - val_accuracy: 0.9319\n", + "Epoch 93/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2524 - accuracy: 0.8974\n", + "Epoch 00093: val_accuracy did not improve from 0.94062\n", + "7878/7878 [==============================] - 5s 596us/sample - loss: 0.2531 - accuracy: 0.8973 - val_loss: 0.2004 - val_accuracy: 0.9220\n", + "Epoch 94/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2521 - accuracy: 0.8987\n", + "Epoch 00094: val_accuracy improved from 0.94062 to 0.94138, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 599us/sample - loss: 0.2525 - accuracy: 0.8986 - val_loss: 0.1625 - val_accuracy: 0.9414\n", + "Epoch 95/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.2452 - accuracy: 0.9022\n", + "Epoch 00095: val_accuracy did not improve from 0.94138\n", + "7878/7878 [==============================] - 5s 618us/sample - loss: 0.2448 - accuracy: 0.9023 - val_loss: 0.1595 - val_accuracy: 0.9387\n", + "Epoch 96/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.2563 - accuracy: 0.8986\n", + "Epoch 00096: val_accuracy did not improve from 0.94138\n", + "7878/7878 [==============================] - 5s 624us/sample - loss: 0.2564 - accuracy: 0.8985 - val_loss: 0.1554 - val_accuracy: 0.9338\n", + "Epoch 97/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2488 - accuracy: 0.8997\n", + "Epoch 00097: val_accuracy did not improve from 0.94138\n", + "7878/7878 [==============================] - 5s 601us/sample - loss: 0.2489 - accuracy: 0.8996 - val_loss: 0.1550 - val_accuracy: 0.9414\n", + "Epoch 98/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.2528 - accuracy: 0.8949\n", + "Epoch 00098: val_accuracy improved from 0.94138 to 0.94328, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 610us/sample - loss: 0.2532 - accuracy: 0.8950 - val_loss: 0.1612 - val_accuracy: 0.9433\n", + "Epoch 99/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2539 - accuracy: 0.8945\n", + "Epoch 00099: val_accuracy improved from 0.94328 to 0.94442, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 601us/sample - loss: 0.2548 - accuracy: 0.8944 - val_loss: 0.1514 - val_accuracy: 0.9444\n", + "Epoch 100/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2589 - accuracy: 0.8952\n", + "Epoch 00100: val_accuracy did not improve from 0.94442\n", + "7878/7878 [==============================] - 5s 603us/sample - loss: 0.2578 - accuracy: 0.8959 - val_loss: 0.1544 - val_accuracy: 0.9387\n", + "Epoch 101/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2405 - accuracy: 0.9027\n", + "Epoch 00101: val_accuracy did not improve from 0.94442\n", + "7878/7878 [==============================] - 5s 595us/sample - loss: 0.2408 - accuracy: 0.9025 - val_loss: 0.1719 - val_accuracy: 0.9277\n", + "Epoch 102/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2443 - accuracy: 0.9052\n", + "Epoch 00102: val_accuracy did not improve from 0.94442\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.2441 - accuracy: 0.9052 - val_loss: 0.1811 - val_accuracy: 0.9319\n", + "Epoch 103/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2376 - accuracy: 0.9054\n", + "Epoch 00103: val_accuracy did not improve from 0.94442\n", + "7878/7878 [==============================] - 5s 596us/sample - loss: 0.2380 - accuracy: 0.9053 - val_loss: 0.1633 - val_accuracy: 0.9383\n", + "Epoch 104/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.2372 - accuracy: 0.9056\n", + "Epoch 00104: ReduceLROnPlateau reducing learning rate to 0.0008153727091848849.\n", + "\n", + "Epoch 00104: val_accuracy did not improve from 0.94442\n", + "7878/7878 [==============================] - 5s 591us/sample - loss: 0.2379 - accuracy: 0.9053 - val_loss: 0.1705 - val_accuracy: 0.9383\n", + "Epoch 105/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.2332 - accuracy: 0.9042\n", + "Epoch 00105: val_accuracy did not improve from 0.94442\n", + "7878/7878 [==============================] - 5s 608us/sample - loss: 0.2334 - accuracy: 0.9042 - val_loss: 0.1649 - val_accuracy: 0.9349\n", + "Epoch 106/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2291 - accuracy: 0.9051\n", + "Epoch 00106: val_accuracy improved from 0.94442 to 0.95204, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 598us/sample - loss: 0.2289 - accuracy: 0.9049 - val_loss: 0.1436 - val_accuracy: 0.9520\n", + "Epoch 107/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2397 - accuracy: 0.9013\n", + "Epoch 00107: val_accuracy did not improve from 0.95204\n", + "7878/7878 [==============================] - 5s 594us/sample - loss: 0.2392 - accuracy: 0.9015 - val_loss: 0.1432 - val_accuracy: 0.9440\n", + "Epoch 108/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.2343 - accuracy: 0.9033\n", + "Epoch 00108: val_accuracy did not improve from 0.95204\n", + "7878/7878 [==============================] - 5s 623us/sample - loss: 0.2342 - accuracy: 0.9034 - val_loss: 0.1462 - val_accuracy: 0.9498\n", + "Epoch 109/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2230 - accuracy: 0.9082\n", + "Epoch 00109: val_accuracy improved from 0.95204 to 0.95775, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 619us/sample - loss: 0.2234 - accuracy: 0.9081 - val_loss: 0.1333 - val_accuracy: 0.9577\n", + "Epoch 110/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2272 - accuracy: 0.9106\n", + "Epoch 00110: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 602us/sample - loss: 0.2269 - accuracy: 0.9103 - val_loss: 0.1502 - val_accuracy: 0.9437\n", + "Epoch 111/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2245 - accuracy: 0.9084\n", + "Epoch 00111: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 599us/sample - loss: 0.2248 - accuracy: 0.9076 - val_loss: 0.1516 - val_accuracy: 0.9410\n", + "Epoch 112/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2170 - accuracy: 0.9088\n", + "Epoch 00112: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 648us/sample - loss: 0.2164 - accuracy: 0.9090 - val_loss: 0.1702 - val_accuracy: 0.9391\n", + "Epoch 113/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2317 - accuracy: 0.9060\n", + "Epoch 00113: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 659us/sample - loss: 0.2315 - accuracy: 0.9063 - val_loss: 0.1808 - val_accuracy: 0.9269\n", + "Epoch 114/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2235 - accuracy: 0.9087\n", + "Epoch 00114: ReduceLROnPlateau reducing learning rate to 0.0007827577739953995.\n", + "\n", + "Epoch 00114: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 626us/sample - loss: 0.2243 - accuracy: 0.9085 - val_loss: 0.1606 - val_accuracy: 0.9360\n", + "Epoch 115/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.2114 - accuracy: 0.9114\n", + "Epoch 00115: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.2125 - accuracy: 0.9114 - val_loss: 0.1377 - val_accuracy: 0.9528\n", + "Epoch 116/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.2252 - accuracy: 0.9056\n", + "Epoch 00116: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 601us/sample - loss: 0.2251 - accuracy: 0.9057 - val_loss: 0.1392 - val_accuracy: 0.9452\n", + "Epoch 117/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2184 - accuracy: 0.9107\n", + "Epoch 00117: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.2182 - accuracy: 0.9108 - val_loss: 0.1342 - val_accuracy: 0.9562\n", + "Epoch 118/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.2171 - accuracy: 0.9114\n", + "Epoch 00118: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 609us/sample - loss: 0.2172 - accuracy: 0.9115 - val_loss: 0.1452 - val_accuracy: 0.9498\n", + "Epoch 119/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2178 - accuracy: 0.9132\n", + "Epoch 00119: ReduceLROnPlateau reducing learning rate to 0.0007514474540948867.\n", + "\n", + "Epoch 00119: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 595us/sample - loss: 0.2180 - accuracy: 0.9133 - val_loss: 0.1441 - val_accuracy: 0.9490\n", + "Epoch 120/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.2089 - accuracy: 0.9122\n", + "Epoch 00120: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 621us/sample - loss: 0.2088 - accuracy: 0.9125 - val_loss: 0.1427 - val_accuracy: 0.9482\n", + "Epoch 121/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.2008 - accuracy: 0.9188\n", + "Epoch 00121: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 624us/sample - loss: 0.2006 - accuracy: 0.9189 - val_loss: 0.1476 - val_accuracy: 0.9418\n", + "Epoch 122/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.2197 - accuracy: 0.9137\n", + "Epoch 00122: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 607us/sample - loss: 0.2197 - accuracy: 0.9138 - val_loss: 0.1375 - val_accuracy: 0.9501\n", + "Epoch 123/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2025 - accuracy: 0.9160\n", + "Epoch 00123: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 600us/sample - loss: 0.2033 - accuracy: 0.9160 - val_loss: 0.1282 - val_accuracy: 0.9505\n", + "Epoch 124/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2036 - accuracy: 0.9132\n", + "Epoch 00124: ReduceLROnPlateau reducing learning rate to 0.0007213895581662655.\n", + "\n", + "Epoch 00124: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 597us/sample - loss: 0.2040 - accuracy: 0.9130 - val_loss: 0.1450 - val_accuracy: 0.9536\n", + "Epoch 125/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2020 - accuracy: 0.9180\n", + "Epoch 00125: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 607us/sample - loss: 0.2014 - accuracy: 0.9183 - val_loss: 0.1348 - val_accuracy: 0.9539\n", + "Epoch 126/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1980 - accuracy: 0.9232\n", + "Epoch 00126: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 598us/sample - loss: 0.1977 - accuracy: 0.9235 - val_loss: 0.1331 - val_accuracy: 0.9539\n", + "Epoch 127/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.2011 - accuracy: 0.9188\n", + "Epoch 00127: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.2010 - accuracy: 0.9185 - val_loss: 0.1197 - val_accuracy: 0.9551\n", + "Epoch 128/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.2006 - accuracy: 0.9182\n", + "Epoch 00128: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 608us/sample - loss: 0.2004 - accuracy: 0.9183 - val_loss: 0.1427 - val_accuracy: 0.9467\n", + "Epoch 129/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1979 - accuracy: 0.9215\n", + "Epoch 00129: ReduceLROnPlateau reducing learning rate to 0.0006925339624285697.\n", + "\n", + "Epoch 00129: val_accuracy did not improve from 0.95775\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.1967 - accuracy: 0.9221 - val_loss: 0.1423 - val_accuracy: 0.9505\n", + "Epoch 130/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1912 - accuracy: 0.9251\n", + "Epoch 00130: val_accuracy improved from 0.95775 to 0.95889, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 604us/sample - loss: 0.1913 - accuracy: 0.9251 - val_loss: 0.1290 - val_accuracy: 0.9589\n", + "Epoch 131/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1904 - accuracy: 0.9244\n", + "Epoch 00131: val_accuracy improved from 0.95889 to 0.95965, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 606us/sample - loss: 0.1908 - accuracy: 0.9242 - val_loss: 0.1197 - val_accuracy: 0.9596\n", + "Epoch 132/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1963 - accuracy: 0.9216\n", + "Epoch 00132: val_accuracy did not improve from 0.95965\n", + "7878/7878 [==============================] - 5s 604us/sample - loss: 0.1963 - accuracy: 0.9217 - val_loss: 0.1512 - val_accuracy: 0.9421\n", + "Epoch 133/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.1851 - accuracy: 0.9249\n", + "Epoch 00133: val_accuracy did not improve from 0.95965\n", + "7878/7878 [==============================] - 5s 636us/sample - loss: 0.1859 - accuracy: 0.9243 - val_loss: 0.1557 - val_accuracy: 0.9376\n", + "Epoch 134/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.2019 - accuracy: 0.9201\n", + "Epoch 00134: val_accuracy did not improve from 0.95965\n", + "7878/7878 [==============================] - 5s 622us/sample - loss: 0.2019 - accuracy: 0.9200 - val_loss: 0.1225 - val_accuracy: 0.9528\n", + "Epoch 135/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.1958 - accuracy: 0.9182\n", + "Epoch 00135: val_accuracy did not improve from 0.95965\n", + "7878/7878 [==============================] - 5s 607us/sample - loss: 0.1963 - accuracy: 0.9180 - val_loss: 0.1324 - val_accuracy: 0.9558\n", + "Epoch 136/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1999 - accuracy: 0.9165\n", + "Epoch 00136: ReduceLROnPlateau reducing learning rate to 0.0006648326106369495.\n", + "\n", + "Epoch 00136: val_accuracy did not improve from 0.95965\n", + "7878/7878 [==============================] - 5s 615us/sample - loss: 0.1994 - accuracy: 0.9167 - val_loss: 0.1193 - val_accuracy: 0.9566\n", + "Epoch 137/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1890 - accuracy: 0.9185\n", + "Epoch 00137: val_accuracy did not improve from 0.95965\n", + "7878/7878 [==============================] - 5s 622us/sample - loss: 0.1889 - accuracy: 0.9186 - val_loss: 0.1228 - val_accuracy: 0.9570\n", + "Epoch 138/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.1946 - accuracy: 0.9217\n", + "Epoch 00138: val_accuracy did not improve from 0.95965\n", + "7878/7878 [==============================] - 5s 605us/sample - loss: 0.1945 - accuracy: 0.9218 - val_loss: 0.1398 - val_accuracy: 0.9494\n", + "Epoch 139/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1860 - accuracy: 0.9232\n", + "Epoch 00139: val_accuracy improved from 0.95965 to 0.96346, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 604us/sample - loss: 0.1860 - accuracy: 0.9231 - val_loss: 0.1213 - val_accuracy: 0.9635\n", + "Epoch 140/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.1934 - accuracy: 0.9226\n", + "Epoch 00140: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.1927 - accuracy: 0.9229 - val_loss: 0.1513 - val_accuracy: 0.9410\n", + "Epoch 141/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.1966 - accuracy: 0.9183\n", + "Epoch 00141: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 593us/sample - loss: 0.1965 - accuracy: 0.9184 - val_loss: 0.1348 - val_accuracy: 0.9494\n", + "Epoch 142/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.1834 - accuracy: 0.9260\n", + "Epoch 00142: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 596us/sample - loss: 0.1834 - accuracy: 0.9259 - val_loss: 0.1308 - val_accuracy: 0.9463\n", + "Epoch 143/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1827 - accuracy: 0.9270\n", + "Epoch 00143: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 591us/sample - loss: 0.1837 - accuracy: 0.9265 - val_loss: 0.1355 - val_accuracy: 0.9555\n", + "Epoch 144/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1905 - accuracy: 0.9206\n", + "Epoch 00144: ReduceLROnPlateau reducing learning rate to 0.0006382392905652523.\n", + "\n", + "Epoch 00144: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 594us/sample - loss: 0.1904 - accuracy: 0.9209 - val_loss: 0.1226 - val_accuracy: 0.9574\n", + "Epoch 145/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.1773 - accuracy: 0.9268\n", + "Epoch 00145: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 624us/sample - loss: 0.1772 - accuracy: 0.9269 - val_loss: 0.1218 - val_accuracy: 0.9570\n", + "Epoch 146/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.1860 - accuracy: 0.9211\n", + "Epoch 00146: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 612us/sample - loss: 0.1858 - accuracy: 0.9212 - val_loss: 0.1157 - val_accuracy: 0.9608\n", + "Epoch 147/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.1746 - accuracy: 0.9262\n", + "Epoch 00147: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 610us/sample - loss: 0.1745 - accuracy: 0.9263 - val_loss: 0.1244 - val_accuracy: 0.9600\n", + "Epoch 148/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1801 - accuracy: 0.9251\n", + "Epoch 00148: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 596us/sample - loss: 0.1802 - accuracy: 0.9254 - val_loss: 0.1114 - val_accuracy: 0.9612\n", + "Epoch 149/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.1770 - accuracy: 0.9293\n", + "Epoch 00149: ReduceLROnPlateau reducing learning rate to 0.0006127097457647323.\n", + "\n", + "Epoch 00149: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 597us/sample - loss: 0.1768 - accuracy: 0.9294 - val_loss: 0.1253 - val_accuracy: 0.9520\n", + "Epoch 150/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1780 - accuracy: 0.9282\n", + "Epoch 00150: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.1775 - accuracy: 0.9283 - val_loss: 0.1159 - val_accuracy: 0.9596\n", + "Epoch 151/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1844 - accuracy: 0.9262\n", + "Epoch 00151: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 598us/sample - loss: 0.1836 - accuracy: 0.9266 - val_loss: 0.1151 - val_accuracy: 0.9604\n", + "Epoch 152/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1752 - accuracy: 0.9290\n", + "Epoch 00152: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 602us/sample - loss: 0.1755 - accuracy: 0.9293 - val_loss: 0.1211 - val_accuracy: 0.9570\n", + "Epoch 153/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.1682 - accuracy: 0.9329\n", + "Epoch 00153: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 598us/sample - loss: 0.1683 - accuracy: 0.9329 - val_loss: 0.1247 - val_accuracy: 0.9539\n", + "Epoch 154/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1676 - accuracy: 0.9346\n", + "Epoch 00154: ReduceLROnPlateau reducing learning rate to 0.0005882013402879238.\n", + "\n", + "Epoch 00154: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.1676 - accuracy: 0.9342 - val_loss: 0.1189 - val_accuracy: 0.9612\n", + "Epoch 155/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.1664 - accuracy: 0.9323\n", + "Epoch 00155: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 593us/sample - loss: 0.1663 - accuracy: 0.9323 - val_loss: 0.1212 - val_accuracy: 0.9532\n", + "Epoch 156/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.1598 - accuracy: 0.9353\n", + "Epoch 00156: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 596us/sample - loss: 0.1597 - accuracy: 0.9354 - val_loss: 0.1161 - val_accuracy: 0.9608\n", + "Epoch 157/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1638 - accuracy: 0.9278\n", + "Epoch 00157: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 594us/sample - loss: 0.1637 - accuracy: 0.9279 - val_loss: 0.1222 - val_accuracy: 0.9536\n", + "Epoch 158/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1626 - accuracy: 0.9337\n", + "Epoch 00158: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 625us/sample - loss: 0.1621 - accuracy: 0.9337 - val_loss: 0.1224 - val_accuracy: 0.9612\n", + "Epoch 159/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.1792 - accuracy: 0.9259\n", + "Epoch 00159: ReduceLROnPlateau reducing learning rate to 0.0005646732822060585.\n", + "\n", + "Epoch 00159: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 634us/sample - loss: 0.1787 - accuracy: 0.9261 - val_loss: 0.1233 - val_accuracy: 0.9600\n", + "Epoch 160/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1648 - accuracy: 0.9330\n", + "Epoch 00160: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 608us/sample - loss: 0.1654 - accuracy: 0.9330 - val_loss: 0.1219 - val_accuracy: 0.9543\n", + "Epoch 161/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1683 - accuracy: 0.9292\n", + "Epoch 00161: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.1677 - accuracy: 0.9294 - val_loss: 0.1150 - val_accuracy: 0.9570\n", + "Epoch 162/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1588 - accuracy: 0.9349\n", + "Epoch 00162: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.1585 - accuracy: 0.9350 - val_loss: 0.1130 - val_accuracy: 0.9574\n", + "Epoch 163/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1658 - accuracy: 0.9305\n", + "Epoch 00163: val_accuracy did not improve from 0.96346\n", + "7878/7878 [==============================] - 5s 594us/sample - loss: 0.1659 - accuracy: 0.9304 - val_loss: 0.1063 - val_accuracy: 0.9619\n", + "Epoch 164/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1657 - accuracy: 0.9346\n", + "Epoch 00164: val_accuracy improved from 0.96346 to 0.96422, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 598us/sample - loss: 0.1651 - accuracy: 0.9345 - val_loss: 0.1090 - val_accuracy: 0.9642\n", + "Epoch 165/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.1674 - accuracy: 0.9314\n", + "Epoch 00165: val_accuracy did not improve from 0.96422\n", + "7878/7878 [==============================] - 5s 601us/sample - loss: 0.1675 - accuracy: 0.9312 - val_loss: 0.1101 - val_accuracy: 0.9574\n", + "Epoch 166/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.1666 - accuracy: 0.9300\n", + "Epoch 00166: val_accuracy did not improve from 0.96422\n", + "7878/7878 [==============================] - 5s 602us/sample - loss: 0.1666 - accuracy: 0.9301 - val_loss: 0.1074 - val_accuracy: 0.9638\n", + "Epoch 167/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1626 - accuracy: 0.9348\n", + "Epoch 00167: val_accuracy did not improve from 0.96422\n", + "7878/7878 [==============================] - 5s 597us/sample - loss: 0.1622 - accuracy: 0.9351 - val_loss: 0.1098 - val_accuracy: 0.9551\n", + "Epoch 168/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1708 - accuracy: 0.9284\n", + "Epoch 00168: val_accuracy improved from 0.96422 to 0.96460, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 604us/sample - loss: 0.1707 - accuracy: 0.9287 - val_loss: 0.1072 - val_accuracy: 0.9646\n", + "Epoch 169/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1678 - accuracy: 0.9315\n", + "Epoch 00169: val_accuracy did not improve from 0.96460\n", + "7878/7878 [==============================] - 5s 593us/sample - loss: 0.1676 - accuracy: 0.9315 - val_loss: 0.1171 - val_accuracy: 0.9612\n", + "Epoch 170/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1642 - accuracy: 0.9335\n", + "Epoch 00170: val_accuracy did not improve from 0.96460\n", + "7878/7878 [==============================] - 5s 599us/sample - loss: 0.1645 - accuracy: 0.9334 - val_loss: 0.1172 - val_accuracy: 0.9596\n", + "Epoch 171/180\n", + "7840/7878 [============================>.] - ETA: 0s - loss: 0.1589 - accuracy: 0.9349\n", + "Epoch 00171: val_accuracy did not improve from 0.96460\n", + "7878/7878 [==============================] - 5s 629us/sample - loss: 0.1586 - accuracy: 0.9350 - val_loss: 0.1035 - val_accuracy: 0.9635\n", + "Epoch 172/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.1577 - accuracy: 0.9353\n", + "Epoch 00172: val_accuracy did not improve from 0.96460\n", + "7878/7878 [==============================] - 5s 632us/sample - loss: 0.1576 - accuracy: 0.9354 - val_loss: 0.1117 - val_accuracy: 0.9596\n", + "Epoch 173/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1559 - accuracy: 0.9338\n", + "Epoch 00173: ReduceLROnPlateau reducing learning rate to 0.0005420863442122937.\n", + "\n", + "Epoch 00173: val_accuracy did not improve from 0.96460\n", + "7878/7878 [==============================] - 5s 597us/sample - loss: 0.1556 - accuracy: 0.9340 - val_loss: 0.1101 - val_accuracy: 0.9558\n", + "Epoch 174/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1587 - accuracy: 0.9357\n", + "Epoch 00174: val_accuracy did not improve from 0.96460\n", + "7878/7878 [==============================] - 5s 597us/sample - loss: 0.1587 - accuracy: 0.9355 - val_loss: 0.1038 - val_accuracy: 0.9623\n", + "Epoch 175/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1516 - accuracy: 0.9392\n", + "Epoch 00175: val_accuracy did not improve from 0.96460\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.1520 - accuracy: 0.9388 - val_loss: 0.1086 - val_accuracy: 0.9536\n", + "Epoch 176/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1456 - accuracy: 0.9410\n", + "Epoch 00176: val_accuracy did not improve from 0.96460\n", + "7878/7878 [==============================] - 5s 593us/sample - loss: 0.1454 - accuracy: 0.9411 - val_loss: 0.1097 - val_accuracy: 0.9596\n", + "Epoch 177/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1529 - accuracy: 0.9378\n", + "Epoch 00177: val_accuracy did not improve from 0.96460\n", + "7878/7878 [==============================] - 5s 622us/sample - loss: 0.1538 - accuracy: 0.9372 - val_loss: 0.1149 - val_accuracy: 0.9581\n", + "Epoch 178/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1510 - accuracy: 0.9362\n", + "Epoch 00178: ReduceLROnPlateau reducing learning rate to 0.0005204028636217117.\n", + "\n", + "Epoch 00178: val_accuracy did not improve from 0.96460\n", + "7878/7878 [==============================] - 5s 655us/sample - loss: 0.1506 - accuracy: 0.9364 - val_loss: 0.1096 - val_accuracy: 0.9623\n", + "Epoch 179/180\n", + "7872/7878 [============================>.] - ETA: 0s - loss: 0.1500 - accuracy: 0.9334\n", + "Epoch 00179: val_accuracy improved from 0.96460 to 0.96726, saving model to CNN_model_2.h5\n", + "7878/7878 [==============================] - 5s 648us/sample - loss: 0.1499 - accuracy: 0.9335 - val_loss: 0.1008 - val_accuracy: 0.9673\n", + "Epoch 180/180\n", + "7808/7878 [============================>.] - ETA: 0s - loss: 0.1632 - accuracy: 0.9322\n", + "Epoch 00180: val_accuracy did not improve from 0.96726\n", + "7878/7878 [==============================] - 5s 592us/sample - loss: 0.1636 - accuracy: 0.9322 - val_loss: 0.1129 - val_accuracy: 0.9543\n" + ] + } + ], + "source": [ + "#21\n", + "history = model.fit(X_train, y_train, epochs=epoch, \n", + " validation_data=(X_cv, y_cv), \n", + " callbacks=callbacks,\n", + " verbose=1)" + ] + }, + { + "cell_type": "code", + "execution_count": 40, + "metadata": {}, + "outputs": [ + { + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "needs_background": "light" + }, + "output_type": "display_data" + } + ], + "source": [ + "#22\n", + "def plot_acc_loss(history, epochs):\n", + " acc = history.history['accuracy']\n", + " loss = history.history['loss']\n", + " val_acc = history.history['val_accuracy']\n", + " val_loss = history.history['val_loss']\n", + " plt.figure(figsize=(20, 8))\n", + " plt.subplot(121)\n", + " plt.plot(range(1,epochs), acc[1:], label='Train Accuracy')\n", + " plt.plot(range(1,epochs), val_acc[1:], label='Test Accuracy')\n", + " plt.title('Total: ' + str(epochs) + ' runs; '+'ACCURACY', size=18)\n", + " plt.legend()\n", + " plt.grid(True)\n", + " plt.subplot(122)\n", + " plt.plot(range(1,epochs), loss[1:], label='Train Loss')\n", + " plt.plot(range(1,epochs), val_loss[1:], label='Test Loss')\n", + " plt.title('Total: ' + str(epochs) + ' runs; '+'LOSS', size=18)\n", + " plt.legend()\n", + " plt.grid(True)\n", + " plt.show()\n", + " \n", + "plot_acc_loss(history, 180)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "_cell_guid": "79c7e3d0-c299-4dcb-8224-4455121ee9b0", + "_uuid": "d629ff2d2480ee46fbb7e2d37f6b5fab8052498a" + }, + "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": "3.7.4" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +}