is there any way to set forecasting horizon for MLP time series forecasting? After training and test - I can forecast only a single value. Forecatsing with ARIMA has this option but don’t see such options for MLP (number of periods to forecast). Suggestions will be extremely helpful.
why do you want to use an MLP model? If you want to use neural nets, then typically RNN’s are a better fit for sequence data. KNIME provides an easy-to-use implementation of the rich Keras toolkit with different types of RNN’s, that can predict a seq2seq task. If you want to you a fully-connected MLP, then you can keep several neurons instead of a single one in the output layer, but i have never used it myself. In general I would start with a simple linear regression baseline to see if RNN’s are worth the effort.
Thanks Misha, will surely try this out and build a case for RNN if relevant.