Closed inpefess closed 1 year ago
You are correct that the item-sessions map can be derived from the session-items map. The reason why both maps are provided is related to the data preprocessing step, where both maps are derived in a single function to avoid computational repetitions.
In normal circumstances, the user may provide the session-items map only, and I can leave the item-sessions map as an optional parameter. I will add this functionality to the fit()
method.
Ok, it's done. Now user can store only session-items mapping (or raw session events).
The
fit
method expects two maps (sessions2items and items2sessions), but we can always construct the second argument from the first, can't we? What is the purpose of such data duplication? Why can't it be part of thefit
method (and as a consequence, why does one have to store such datasets)? It might be worth mentioning in the JOSS paper (there is only the first argument now mentioned in the draft).