I am building an LSTM model with the following characteristics:
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
lstm (LSTM) (None, 2, 200) 164000
leaky_re_lu (LeakyReLU) (None, 2, 200) 0
lstm_1 (LSTM) (None, 2, 128) 168448
leaky_re_lu_1 (LeakyReLU) (None, 2, 128) 0
dropout (Dropout) (None, 2, 128) 0
lstm_2 (LSTM) (None, 64) 49408
dropout_1 (Dropout) (None, 64) 0
dense (Dense) (None, 1) 65
=================================================================
Total params: 381,921
Trainable params: 381,921
Non-trainable params: 0
_________________________________________________________________
I am a beginner at forecasting and I would like to know if the chosen model is accurate.
These are the results for the plot of loss train and validation loss.
I saw that the loss for train is bigger than the one for validation. Can this be overfitting of the model?
What can be done about it to improve the model?