Generating Top-N Recommendations for the Recommendation System

[PurushothamanSrikanth]

Generating Top-N Recommendations for the Recommendation System

[miguelgfierro]

The right way of doing it is by changing the loss function to a ranking loss (like pair-wise loss).

An easier path, but not totally correct but that many people do, is to get the results the predictions, sort them and get the top-N. It is not correct because the loss is not intended for the task of ranking. But again, it is a technique that many people use because it is easier.

[PurushothamanSrikanth]

The right way of doing it is by changing the loss function to a ranking loss (like pair-wise loss).

An easier path, but not totally correct but that many people do, is to get the results the predictions, sort them and get the top-N. It is not correct because the loss is not intended for the task of ranking. But again, it is a technique that many people use because it is easier.

@miguelgfierro Can we have this Top-N recommendation as a new feature to the existing one? #112
We can generate Top-N recommendations from the rating prediction matrix.

This is a feature request 😇