ImplicitMF is a Python package that generates personalized recommendations for implicit feedback datasets. Unlike explicit feedback (e.g., movie ratings), implicit feedback looks at a user's interactions with an item and uses this as a surrogate measure of their preference toward that item.
ImplicitMF provides a set of tools for building a recommendation system pipeline. These tools facilitate data pre-processing, hyperparameter training, model fitting, evaluation, validation, and post-processing of results. ImplicitMF focuses on two types of matrix factorization models that are built specifically for implicit feedback datasets:
- Alternating Least Squares (ALS): as described in Collaborative Filtering for Implicit Feedback Datasets. See implicit package for more information on its Python implementation.
- Learning to Rank: as described in BPR: Bayesian Personalized Ranking from Implicit Feedback and WSABIE: Scaling Up To Large Vocabulary Image Annotation. See LightFM package for more information on its Python implementation.
Further documentation can be found here.