[2Win5] Poster session 2

[2Win5-28] Analyzing Temporal Changes in User Embeddings in Collaborative Filtering Recommender Systems

〇Kazuo Watanabe¹ (1.The Japan Research Institute, Limited)

Keywords:Recommender System, Collaborative Filtering, Cold-start problem

Collaborative filtering is one of the most popular and successful approaches in recommender systems. However, the cold-start problem is a significant challenge in collaborative filtering. Although there are many studies addressing the cold-start problem, most studies lack a unified definition of cold users and rely on their own experimental settings. In this paper, we observerd the temporal changes in the users' embeddings in collaborative filltering, and quantitatively evaluated the correlation between the number of records and the similarity of users' embeddings.