[Idea] Implicit feedback collaborative filtering

[ankane]

Hi, I think xLearn can be a great tool for recommender systems. It works great for explicit feedback following this approach with both FM and FFM.

I was wondering if you'd consider supporting implicit feedback (one-class matrix factorization). Here are a few papers on the topic:

• FM
  • Factorization Machines for Datasets with Implicit Feedback
  • Optimizing Factorization Machines for Top-N Context-Aware Recommendations
  • BoostFM: Boosted Factorization Machines for Top-N Feature-based Recommendation
• FFM
  • One-class Field-aware Factorization Machines for Recommender Systems with Implicit Feedbacks

Related to #230

[sumitsidana]

Hi @ankane thanks for sharing these resources. I am curious can't the problem of implicit feedback be reduced to binary classification by doing negative sampling and then applying the original FM or FFM implementation?

[ankane]

Hi @sumitsidana, thanks for the response. Unfortunately, I'm not super familiar with the topic, so don't know the answer to that question.

I've noticed some MF (not FM) libraries like libmf use a different approach for implicit feedback than explicit feedback (I believe performance is a factor), so not sure if something like that applies here. If not, maybe there could be built-in support for the approach mentioned above.

Sorry, this is probably one of my least helpful responses on GitHub.