Title: A study on recommender system considering diversity of items based on LDA

Authors: Zhiying Zhang; Taiju Hosaka; Haruka Yamashita; Masayuki Goto

Addresses: Graduate School of Creative Science and Engineering, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, 169-8555, Japan ' Sansan, Inc., Aoyama Oval Building 13F, 5-52-2 Jingumae, Shibuya-ku, Tokyo, 150-0001, Japan ' School of Information and Computer Science, Sophia University, 7-1, Kioi cho, Chiyoda-ku, Tokyo, 102-8854, Japan ' School of Creative Science and Engineering, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, 169-8555, Japan

Abstract: With the rapid development of information technology, a recommender system making use of users' behaviour data, such as browsing history or ratings for items, is now one of the important tools for searching contents or products. Recently, it has been shown that diversifying the recommendation lists in recommender systems could satisfy users' potential needs. In a previous research, the diversity of recommender system can be raised by the topic diversification method based on latent Dirichlet allocation (LDA); however, since the items belonging to the same topic are not diversified, the recommended items in the list shown to a user tend to be similar. Therefore, this research proposes a method for a recommendation system that diversifies items in each topic based on topic information obtained by LDA. Experimental results with MovieLens datasets demonstrate that our approach keeps accuracy of the recommendation and realises more diversified recommendation.

Keywords: recommender system; latent Dirichlet allocation; LDA; topic model; machine learning; diversity.

DOI: 10.1504/AJMSA.2021.118391

Asian Journal of Management Science and Applications, 2021 Vol.6 No.1, pp.17 - 31

Received: 28 Jun 2020
Accepted: 07 Nov 2020
Published online: 25 Oct 2021 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article