Hi. let me start by explaining some lstm Keras definition. suppose we want to predict the next 21 day stock price (21st day not first to 21st day, only 1 day) from the last 30 days history. if we have 2 features (close price and volume), our input dataset will look like [num_sample,30,2] and output [num_sample,1(21st)].
How can I implement this in the gluonTS dataset? we have a start, target and feat_dynamic_real parameters. should I set my output in target and input (num_sample,30*2) in feat_dynamic_real?
or not just set prediction_length to 21, context_length to 30 and feat_dynamic_real to (num_sample,2).
thanks.