I have a question regarding the time series forecasting.
the model is trained in two different way:
- to perform one step forecasting with 60 time step of past values.
- to perform 5 steps ahead forecasting with 60 time step of past values.
My question is whether multiple step ahead forecasting will improve accuracy or not?
I have trained one step forecasting and obtain the MSE of 0.045.
I also trained 5 steps ahead forecasting, but I calculated MSE for the one step ahead forecasting and I got MSE of 0.0047.
I tried to compare the MSE of one step ahead forecast for both model.
My hypothesis is, if we train a model with multiple steps ahead forecast will improve the accuracy of one step forecast. But the MSE didn’t show any improvement.
May I know is my hypothesis correct? Is model trained withmultiple step ahead forecasting will improve the accuracy of the one step ahead accuracy compare to model train with one step ahead forecasting?
Thank you very much