Temporal Positive Collective Matrix Factorization for Interpretable Trend Analysis in Recommender Systems

Matrix Factorization (MF) is a common method in Recommender Systems (RS). However, distinguishing between continuously and temporarily popular items is challenging, as the basic MF relies on the accumulated user rating records for an item. Moreover, recent trends emphasize that RS should not only be accurate but also interpretable, necessitating clear reasons behind each recommendation. We propose the Temporal Positive Collective Matrix Factorization (TPCMF) method, which improves the interpretability and temporality of the Collective Matrix Factorization (CMF). We make the factor matrix obtained by CMF non-negative to increase the interpretability. In addition, we take into account the time variations of the factor matrices and make time series predictions, which enables temporal recommendations. Moreover, in an experiment using a real-world dataset, we determined that factor interpretation under TPCMF provides substantial insights into the interpretability and temporality of recommendations with accuracy surpassing existing methods.