The task is as follows:
- to build in Matlab neural networks forecasting industrial production and retail sales one, two and three months ahead. Output needed is quite standard - forecasts, statistics, performance (errors).
IMPORTANT - The forecasts in one month horizon will have to beat competing econometric models
Errors and data are provided in attached files (for retail sales, for industrial output later).
- to write around 2 pages (possibly in bullets) why this kind of network was used, and not the other one. I need explanations about each choice/decision made during nn construction (like why this number of epochs, why this and not that way etc). It could be in a tree or mindmap way - I need to see how many options there were and why this particular was chosen.
I just would like to generally understand why the nn's are contructed this particular way.
The code have to be written in a user friendly way with explanations.