Forecasting Large Hierarchies

[jrodenbergrheem]

I am forecasting a very large hierarchy and wanted to know anyone's approach on filtering out redundant hierarchy chains, is this a post or preprocessing step? And when doing this is it better to use a particular strategy?