An integration method for optimizing the use of explicit and implicit feedback in recommender systems

초록

The recent changes in consumption patterns and the development of the Internet have increased the diversity of user feedback in the recommender system. As a result, recent studies have highlighted the complementary integration of heterogeneous feedbacks to improve the quality of recommendations. However, existing integration methods tend to be biased toward one type of feedback, which hinders proper integration, and overlook the information loss problem caused by joint training of heterogeneous feedbacks. In this work, a novel method for integrating explicit and implicit feedback (IEIF) is proposed to generate a new user preference for the personalized recommender system. The IEIF complements the data shortage by containing various types of implicit feedback, while maintaining the item ranking results derived from user-provided explicit feedback. Through extensive experiments conducted on three real-world datasets, the superior performances of the IEIF are demonstrated with improvements in precision, recall, and NDCG, with average gains of 9.92%, 8.71%, and 6.04%, respectively, over the other integration methods. © 2023, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.

키워드