Recommendation incorporating transition of temporally intensive unity
Kenta Inuzuka, Tomonori Hayashi, Tomohiro Takagi · 2016
It is important to note that user preferences change over time. However, it is not guaranteed that user preferences change at a steady rate. For example, a person who intensively listens to music of the same artist might intensively listen to the music of a different artist after a few days. For this reason, it is effective to incorporate such preference changes into recommender systems. In this paper, we propose an approach that predicts user preferences with consideration of preference changes by learning the transition of the preference that is the temporally intensive unity of purchasing items as one preference. Our approach is composed of a Kalman filter and matrix factorization. We show through experiments using a real-world dataset that our approach outperforms competitive methods such as the first order Markov model.