validation_data it shows the accuracy whether training is going well or no, it doesn't train validation data. Once loss is turn from goes down to goes up, it is a signal of overfitting. model=Sequential() model.add(Dense(3,input_dim=4, activation='softmax', input_shape=X_train[0].shape)) model.compile(loss='categorical_crossentropy', optimizer='adam',metrics=['accuracy']) history=model.fit(X_tra..