Fusion-based Recommender System for Improving Serendipity

Novelty and Diversity in Recommender Systems 2011.

计算机科学;

Kenta Oku ; Fumio Hattori

Others  :  http://ceur-ws.org/Vol-816/paper3.pdf
PID  :  42542

学科分类：计算机科学（综合）

来源: CEUR

【 摘 要 】

Recent work has focused on new measures that are beyond the accuracy of recommender systems. Serendipity, which is one of these measures, is defined as a measure that indicates how the recommender system can find unexpected and useful items for users. In this paper, we propose a Fusion-based Recommender System that aims to improve the serendipity of recommender systems. The system is based on the novel notion that the system finds new items, which have the mixed features of two user-input items, produced by mixing the two items together. The system consists of item-fusion methods and scoring methods. The item-fusion methods generate a recommendation list based on mixed features of two user-input items. Scoring methods are used to rank the recommendation list. This paper describes these methods and gives experimental results.