Cannot build hierarchy with large dataset about scikit-hts

[carlomazzaferro]

Thanks for reporting. While I don't have a solution for you right now, it is good to know what kind of performance issues might arise when dealing with larger datasets. Just to give some background, the hierarchy tree is build by keeping in memory each node (represented as a pandas series/dataframe (in case you add exogenous variables)).

In the future, it might be possible to defer the instantiation of the series until the fit procedure is called, so I'll add an issue for that use-case.