I have a limited dataset with four timesteps, and each timestep has 17 variables to predict. So the input shape is input_shape=(4, 17). For example for a person with ID 1, the would be like this:
time 1 = x1,x2,x3,x4..., x16, x17
time 2 = x1,x2,x3,x4..., x16, x17
time 3 = x1,x2,x3,x4..., x16, x17
time 4 = x1,x2,x3,x4..., x16, x17
There is 1 Y for each timestep. I need to have the output for each timestep. Here is my code:
model = Sequential()
model.add(Masking(mask_value=-2., input_shape=(4, 17)))
model.add(LSTM(150, activation='tanh', return_sequences=True))
model.add(LSTM(150, activation='tanh', return_sequences=True))
model.add(Dropout(0.2))
model.add(Dense(1, activation='relu'))
model.compile(optimizer='adam', loss='mean_absolute_error')
model.summary()
I have add return_sequence = True for each LSTM layer. But I'm not sure if this is the right code or not.
CodePudding user response：
getting prediction for each timestep in your LSTM has more to do with pre-processing the data than it does the actual design of the LSTM. This video may be able to provide you with some helpful information, and provides a link to the code used to design the LSTM: