{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# Beyond Analysis ","metadata":{}},{"cell_type":"markdown","source":"## Loading required library","metadata":{}},{"cell_type":"code","source":"import pandas as pd\r\nfrom matplotlib import pyplot as plt\r\nimport numpy as np\r\nimport seaborn as sns\r\nimport sklearn\r\n\r\nfrom sklearn.preprocessing import StandardScaler,FunctionTransformer,LabelEncoder,OneHotEncoder,MinMaxScaler\r\nfrom sklearn.model_selection import train_test_split\r\nfrom sklearn.metrics import mean_squared_error\r\nfrom sklearn.compose import ColumnTransformer\r\n\r\nimport tensorflow as tf\r\nfrom tensorflow.keras.layers import LSTM,Dense,Dropout\r\nfrom tensorflow.data import Dataset as tfds\r\n\r\npd.set_option('display.max_columns', None)\r\npd.set_option('display.max_rows', None)\r\n\r\n\r\ndfUIDLen=96298 #for loading processed data directly\r\ndfTestUIDLen=65242","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Loading Data and data preparation","metadata":{}},{"cell_type":"code","source":"df=pd.read_csv(\"train_sort.csv\")\r\ndf.head(30)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"dfTest=pd.read_csv(\"test_sort.csv\",delimiter=',')\r\ndfTest.head(30)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Note:- commented after sorting the given data to reduce computational cost for repeated testing","metadata":{}},{"cell_type":"markdown","source":"df=pd.read_csv('train.csv',delimiter='\\t')\ndf.sort_values('UNIQUE_IDENTIFIER',axis=0,inplace=True)\ndf.reset_index(drop=True,inplace=True)\ndf.to_csv(\"train_sort.csv\",index=False)","metadata":{}},{"cell_type":"markdown","source":"dfTest=pd.read_csv('test.csv')\ndfTest.sort_values('UNIQUE_IDENTIFIER',axis=0,inplace=True)\ndfTest.reset_index(drop=True,inplace=True)\ndfTest.to_csv(\"test_sort.csv\",index=False)","metadata":{}},{"cell_type":"markdown","source":"## Data understanding and transformation","metadata":{}},{"cell_type":"code","source":"dfNoUID=df['UNIQUE_IDENTIFIER'].value_counts()\r\ndfUIDLen=dfNoUID.size","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"dfTestNoUID=dfTest['UNIQUE_IDENTIFIER'].value_counts()\r\ndfTestUIDLen=dfTestNoUID.size","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df.dtypes","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df.shape","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df.describe()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"le1=LabelEncoder()\r\nle2=LabelEncoder()\r\ndfL=df.copy()\r\ndfTestL=dfTest.copy()\r\ndfTestL['CATEGORY_1']=le1.fit_transform(dfTest['CATEGORY_1'])\r\ndfL['CATEGORY_1']=le1.transform(df['CATEGORY_1'])\r\ndfTestL['CATEGORY_2']=le2.fit_transform(dfTest['CATEGORY_2'])\r\ndfL['CATEGORY_2']=le2.transform(df['CATEGORY_2'])","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"corr=dfL.corr()\r\ncorr.style.background_gradient(cmap='coolwarm')","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"dfL=dfL.drop(['CATEGORY_1'],axis=1)\r\ndfTestL=dfTestL.drop(['CATEGORY_1'],axis=1)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"dfL['CATEGORY_2'].value_counts()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#df.isna()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"plt.figure(figsize=(12,8))\r\nax=sns.violinplot(data=dfL.drop(['UNIQUE_IDENTIFIER','Y1','Y2'],axis=1),linewidth=2,orient=\"h\")\r\nax.tick_params(axis='x',rotation=90)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def log2Scale(x):\r\n    return np.log2(x+1)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"ct=ColumnTransformer([\r\n    (\"Le\",OneHotEncoder(),[3]),\r\n    (\"logScale\",FunctionTransformer(log2Scale),[5,6,7,8,9,10,13,15,16,18,19]),\r\n    (\"StandardSca\",StandardScaler(),[11,12,14,17,20,21])],remainder='passthrough')\r\ndfCt=pd.DataFrame(ct.fit_transform(dfL.drop(['Y1','Y2'],axis=1)),columns=np.append(ct.transformers_[0][1].get_feature_names(),dfL.columns[[5,6,7,8,9,10,13,15,16,18,19,11,12,14,17,20,21,0,1,2,4]]))\r\ndfCtTest=pd.DataFrame(ct.fit_transform(dfTestL),columns=np.append(ct.transformers_[0][1].get_feature_names(),dfL.columns[[5,6,7,8,9,10,13,15,16,18,19,11,12,14,17,20,21,0,1,2,4]]))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(dfCt.shape)\r\ndfCt.head()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#sns.boxplot(data=dfCt.drop(['UNIQUE_IDENTIFIER'],axis=1))\r\nplt.figure(figsize=(12,8))\r\nax=sns.violinplot(data=dfCt.drop(['UNIQUE_IDENTIFIER'],axis=1),linewidth=2,orient=\"h\")\r\nax.tick_params(axis='x',rotation=90)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#plotted for all standard scalar normalized data\r\nplt.figure(figsize=(12,8))\r\nax=sns.violinplot(data=dfCt.drop(['UNIQUE_IDENTIFIER'],axis=1),linewidth=2,orient=\"h\")\r\nax.tick_params(axis='x',rotation=90)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(dfCt.shape)\r\nsns.relplot(data=dfL,y='Y2',x='WITHDRAW')","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"daSeqY1Y2=pd.DataFrame(columns={\"ID\",\"Max_Sequence\",\"Y1\",\"Y2\"})\r\ndaSeqY1Y2[\"ID\"]=dfNoUID.keys()\r\ndaSeqY1Y2[\"Max_Sequence\"]=dfNoUID.to_numpy()\r\ny1y2=np.zeros((dfNoUID.size,2))\r\nind=0\r\nfor i in range(0,dfNoUID.size):\r\n    y1y2[i]=dfL[['Y1','Y2']].loc[[ind]].to_numpy()\r\n    ind=ind+daSeqY1Y2[\"Max_Sequence\"].loc[i]\r\n    \r\ndaSeqY1Y2[\"Y1\"]=y1y2[:,0]\r\ndaSeqY1Y2[\"Y2\"]=y1y2[:,1]","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#daSeqY1Y2.to_csv(\"daSeqY1Y2.csv\",index=False)\r\ndaSeqY1Y2=pd.read_csv(\"daSeqY1Y2.csv\")\r\ndaSeqY1Y2.head()\r\n#daSeqY1Y2.dtypes","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sns.relplot(x=daSeqY1Y2[\"Max_Sequence\"],y=daSeqY1Y2['Y1'],kind=\"line\")","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Data preprocessing","metadata":{}},{"cell_type":"code","source":"ind=0\r\nlstm_data_X=np.zeros((dfNoUID.size,30,30))\r\nlstm_data_y=np.zeros((dfNoUID.size,2))\r\nidTrain=np.zeros((dfNoUID.size))\r\nfor i in range(0,dfNoUID.size):\r\n    #idTrain[i]=int(dfCt['UNIQUE_IDENTIFIER'][ind])\r\n    lenTrain=dfNoUID[dfCt['UNIQUE_IDENTIFIER'][ind]]\r\n    lstm_data_X[i,30-lenTrain:30,:]=dfCt.iloc[ind:ind+lenTrain].to_numpy(dtype='float')\r\n    lstm_data_y[i]=dfL[['Y1','Y2']].loc[[ind]].to_numpy()\r\n    ind=ind+lenTrain","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_data_X=np.delete(lstm_data_X,26,axis=2)\r\nlstm_data_X[0]","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"dfCt.iloc[0:7]","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"ind=0\r\nlstm_test_X=np.zeros((dfTestNoUID.size,30,29))\r\nidTest=np.zeros((dfTestNoUID.size))\r\nfor i in range(0,dfTestNoUID.size):\r\n    idTest[i]=int(dfCtTest['UNIQUE_IDENTIFIER'][ind])\r\n    lnth=dfTestNoUID[dfCtTest['UNIQUE_IDENTIFIER'][ind]]\r\n    lstm_test_X[i,30-lnth:30,:]=dfCtTest.iloc[ind:ind+lnth].drop(['UNIQUE_IDENTIFIER'],axis=1).to_numpy(dtype='float')\r\n    ind=ind+lnth","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"1. All standard scalar\n2. All log scaled\n3. Mixed standard scalar+log\n4. Mixed sqrt + log","metadata":{}},{"cell_type":"code","source":"np.savetxt(\"lstm_x_4.csv\",lstm_data_X.reshape(lstm_data_X.shape[0],-1),fmt='%.6e',delimiter=',',newline='\\n')\r\nnp.savetxt(\"lst_y_4.csv\",lstm_data_y,fmt='%.6e',delimiter=',',newline='\\n')\r\nnp.savetxt(\"lstm_test_x_4.csv\",lstm_test_X.reshape(lstm_test_X.shape[0],-1),fmt='%.6e',delimiter=',',newline='\\n')","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Model creating and compilling ","metadata":{}},{"cell_type":"code","source":"lstm_modely1=tf.keras.models.Sequential()\r\nlstm_modely1.add(LSTM(10,activation='sigmoid',input_shape=(30,29),kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\nlstm_modely1.add(Dropout(0.05))\r\nlstm_modely1.add(Dense(10,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\nlstm_modely1.add(Dropout(0.05))\r\nlstm_modely1.add(Dense(5,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\nlstm_modely1.add(Dense(1,activation='relu',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n\r\nlstm_modely1.compile(optimizer='Adam',loss='mse',metrics=[\"mse\",\"mae\"])","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_modely1.summary()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_modely2=tf.keras.models.Sequential()\r\n#lstm_modely2.add(Dropout(0.05,input_shape=(30,29)))\r\nlstm_modely2.add(LSTM(10,activation='sigmoid',input_shape=(30,29),kernel_regularizer=tf.keras.regularizers.L2(0.01)))\r\nlstm_modely2.add(Dense(10,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L2(0.01)))\r\nlstm_modely2.add(Dense(5,activation='relu',kernel_regularizer=tf.keras.regularizers.L2(0.01)))\r\nlstm_modely2.add(Dense(1,activation='relu',kernel_regularizer=tf.keras.regularizers.L2(0.01)))\r\n\r\nlstm_modely2.compile(optimizer='Adam',loss='mse',metrics=[\"mse\",\"mae\"])","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_modely2.summary()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_data_X=np.loadtxt(\"lstm_x_3.csv\",delimiter=',')\r\nlstm_data_X=lstm_data_X.reshape(dfUIDLen,30,29)\r\nlstm_data_y=np.loadtxt(\"lst_y_3.csv\",delimiter=',')\r\nlstm_test_X=np.loadtxt(\"lstm_test_x_3.csv\",delimiter=',')\r\nlstm_test_X=lstm_test_X.reshape(dfTestUIDLen,30,29)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(lstm_data_X.shape,'\\n',lstm_test_X.shape)\r\nt=pd.DataFrame(lstm_data_X[0])\r\nt.head(30)\r\n#lstm_data_y[0]\r\n#lstm_model(lstm_data_X[0:1,:,:]).shape","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_x,test_x,train_y,test_y=train_test_split(lstm_data_X,lstm_data_y,test_size=0.3,random_state=0)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Cross Validation with 5 fold","metadata":{}},{"cell_type":"code","source":"TestXsp1,TestXsp2,TestXsp3,TestXsp4,TestXsp5=tf.split(lstm_data_X[:-3,:,:],5,axis=0,name='split')\r\nTestYsp1,TestYsp2,TestYsp3,TestYsp4,TestYsp5=tf.split(lstm_data_y[:-3,:],5,axis=0,name='split')","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_x=np.zeros((19259*4,30,29))\r\ntrain_y=np.zeros((19259*4,2))\r\ntrain_x=np.concatenate((TestXsp1,TestXsp2,TestXsp3,TestXsp4),axis=0,out=train_x)\r\ntrain_y=np.concatenate((TestYsp1,TestYsp2,TestYsp3,TestYsp4),axis=0,out=train_y)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(train_x.shape,train_y.shape,TestXsp2.shape,TestYsp2.shape)\r\ntrain_y[0]","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_modely1.load_weights('lstmM1y1LSq1.hd5')\r\nlstm_modely2.load_weights('lstmM1y2LSq1.hd5')","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_modely1.fit(train_x,train_y[:,0],batch_size=75,epochs=20)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_modely2.fit(train_x,train_y[:,1],batch_size=75,epochs=20)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_modely1.evaluate(TestXsp5,TestYsp5[:,0])","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_modely2.evaluate(TestXsp5,TestYsp5[:,1])","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#lstm_modely1.save_weights('lstmM1y1LSq1.hd5',overwrite=True,save_format=None,options=None)\r\nlstm_modely1.load_weights('lstmM1y1LS1.hd5')","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#lstm_modely2.save_weights('lstmM1y2LSq1.hd5',overwrite=True,save_format=None,options=None)\r\nlstm_modely2.load_weights('lstmM1y2LS1.hd5')","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"TestXsp1,TestXsp2,TestXsp3=tf.split(test_x,3,axis=0,name='split')\r\nTestYsp1,TestYsp2,TestYsp3=tf.split(test_y,3,axis=0,name='split')","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_modely1.fit(test_x,test_y[:,0],epochs=5)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_modely2.fit(test_x,test_y[:,1],batch_size=25,epochs=5)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_modely1.fit(lstm_data_X,lstm_data_y[:,0],batch_size=75,epochs=2)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"lstm_modely2.fit(lstm_data_X,lstm_data_y[:,1],batch_size=75,epochs=2)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"pred_y1=lstm_modely1.predict(lstm_test_X)\r\npred_y2=lstm_modely2.predict(lstm_test_X)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"pred=pd.read_csv('Sub2.csv')\r\n#pred=pd.DataFrame(columns=['UNIQUE_IDENTIFIER','Y1','Y2'])\r\n#pred.astype({'UNIQUE_IDENTIFIER':'Int64'})\r\n#idTest=idTest.astype('int64')\r\n#pred['UNIQUE_IDENTIFIER']=idTest\r\npred['Y1']=pred_y1\r\n#pred['Y2']=pred_y2","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"pred.to_csv('Sub3.csv',index=False)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#pred.astype({'UNIQUE_IDENTIFIER':'int64'})\r\npred.dtypes","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"pred.head()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Other models for comparasion","metadata":{}},{"cell_type":"code","source":"simple_modely1=tf.keras.models.Sequential()\r\nsimple_modely1.add(Dense(10,activation='sigmoid',input_shape=(29,),kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n#simple_modely1.add(Dropout(0.05))\r\nsimple_modely1.add(Dense(10,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n#simple_modely1.add(Dropout(0.05))\r\nsimple_modely1.add(Dense(5,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\nsimple_modely1.add(Dense(1,activation='relu',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n\r\nsimple_modely1.compile(optimizer='Adam',loss='mse',metrics=[\"mse\",\"mae\"])","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"simple_modely1.summary()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"simple_modely2=tf.keras.models.Sequential()\r\nsimple_modely2.add(Dense(10,activation='sigmoid',input_shape=(29,),kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n#simple_modely1.add(Dropout(0.05))\r\nsimple_modely2.add(Dense(10,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n#simple_modely1.add(Dropout(0.05))\r\nsimple_modely2.add(Dense(5,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\nsimple_modely2.add(Dense(1,activation='relu',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n\r\nsimple_modely2.compile(optimizer='Adam',loss='mse',metrics=[\"mse\",\"mae\"])","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"simple_modely2.summary()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_x,test_x,train_y1,test_y1,train_y2,test_y2=train_test_split(dfCt.drop(['UNIQUE_IDENTIFIER'],axis=1),dfL['Y1'],dfL['Y2'],test_size=0.3,random_state=0)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(train_x.shape,train_y1.shape,test_x.shape,test_y1.shape,train_y2.shape,test_y2.shape)\r\ntrain_y1[0]","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"simple_modely1.fit(train_x,train_y1,batch_size=75,epochs=20)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"simple_modely2.fit(train_x,train_y2,batch_size=75,epochs=10)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"simple_modely1.evaluate(test_x,test_y1)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"simple_modely2.evaluate(test_x,test_y2)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"deep_modely1=tf.keras.models.Sequential()\r\ndeep_modely1.add(Dense(20,activation='sigmoid',input_shape=(29,),kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n#simple_modely1.add(Dropout(0.05))\r\ndeep_modely1.add(Dense(20,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\ndeep_modely1.add(Dense(20,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n#simple_modely1.add(Dropout(0.05))\r\ndeep_modely1.add(Dense(10,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\ndeep_modely1.add(Dense(5,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\ndeep_modely1.add(Dense(1,activation='relu',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n\r\ndeep_modely1.compile(optimizer='Adam',loss='mse',metrics=[\"mse\",\"mae\"])","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"deep_modely1.summary()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"deep_modely2=tf.keras.models.Sequential()\r\ndeep_modely2.add(Dense(20,activation='sigmoid',input_shape=(29,),kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n#simple_modely1.add(Dropout(0.05))\r\ndeep_modely2.add(Dense(20,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\ndeep_modely2.add(Dense(20,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\ndeep_modely2.add(Dense(10,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n#simple_modely1.add(Dropout(0.05))\r\ndeep_modely2.add(Dense(5,activation='sigmoid',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\ndeep_modely2.add(Dense(1,activation='relu',kernel_regularizer=tf.keras.regularizers.L1L2(l1=0.01,l2=0.01)))\r\n\r\ndeep_modely2.compile(optimizer='Adam',loss='mse',metrics=[\"mse\",\"mae\"])","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"deep_modely2.summary()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"deep_modely1.fit(train_x,train_y1,batch_size=75,epochs=20)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"deep_modely2.fit(train_x,train_y2,batch_size=75,epochs=20)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"deep_modely1.evaluate(test_x,test_y1)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"deep_modely2.evaluate(test_x,test_y2)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}