better cnn, 81.4% acc, but confusion matrix isnt great)

python/w2v_cnn_gen_hdf5.py

@@ -10,16 +10,16 @@ from tensorflow.keras.layers import Conv1D,MaxPooling1D,GlobalMaxPooling1D
 from tensorflow import keras
 
 modelNN = Sequential()
-#input_shape=((72, 100)))
 
-modelNN.add(Conv1D(50,kernel_size=5, activation='relu',input_shape=((72, 100))))
+modelNN.add(Conv1D(150,kernel_size=5, activation='relu',input_shape=((72, 100))))
 #modelNN.add(MaxPooling1D(pool_size=4))
 #modelNN.add(Conv1D(250,kernel_size=4, activation='relu'))
 modelNN.add(MaxPooling1D(pool_size=4))
-modelNN.add(Conv1D(25,kernel_size=3, activation='relu'))
+modelNN.add(Conv1D(100,kernel_size=3, activation='relu'))
 modelNN.add(MaxPooling1D(pool_size=4))
 modelNN.add(Flatten())
-modelNN.add(Dense(25,activation='relu'))
+modelNN.add(Dense(300,activation='relu'))
+modelNN.add(Dense(100,activation='relu'))
 #modelNN.add(Dense(50,activation='relu'))
 modelNN.add(Dense(10,activation='relu'))
 modelNN.add(Dense(3,activation='softmax'))
@@ -30,7 +30,7 @@ from hdf5 import hdf5Generator
 path = "G:\\ml\\"
 num_rows = 4.8E6 
 #num_rows = 1E5 
-batchSize = 2048
+batchSize = 256
 steps = num_rows/batchSize
 #early stop
 earlystop = keras.callbacks.EarlyStopping(monitor='val_sparse_categorical_accuracy',patience=5,verbose=False,restore_best_weights=True)
@@ -41,8 +41,8 @@ valData = hdf5Generator(path + "w2vCNN.hdf5", batchSize, "Val")
 
 #%%
 cW = {0:4.18,1:9.53,2:1.52}
-hist = modelNN.fit(trainData, validation_data=valData, epochs=100,class_weight=cW, steps_per_epoch=steps, validation_steps=steps,callbacks=cbList)
-modelNN.save("D:\\ml\\CNN-Classfication-5")
+hist = modelNN.fit(trainData, validation_data=valData, epochs=100,class_weight=cW, steps_per_epoch=steps, validation_steps=int(steps/3),callbacks=cbList)
+modelNN.save("D:\\ml\\CNN-Classfication-6")
 #modelNN.fit(train,epochs=12,validation_data=val,batch_size=batchSize,steps_per_epoch= num_rows/batchSize,callbacks=cbList,validation_steps=num_rows/batchSize)
 # %%eval
 testData = hdf5Generator(path + "w2vCNN.hdf5", batchSize, "Test",loop=False)