Topic Model-Based Recommender System for Longtailed Products Against Popularity Bias
Chunhua Sun, Yinjie Xu · 2019
Future businesses will obtain great profit from long-tailed selling in the reason of meeting the personalized needs of users, however the data sparsity makes long-tailed items hard to be recommended, resulting the problem of popularity bias. To improve the quality of recommendations for long-tailed products, we propose a new topic model-based approach named joint user and social tags model (JUST) to solve the problem, jointly modeling user behavior data, product label information and social relationship. Let labels and social relationship data compensate for the sparsity of the behavior. To assess the effectiveness of our model, we conduct empirical evaluations on the last.fm-2k dataset released by HetRec2011. The experimental results show that the JUST model is better than the compared models in recommending long-tailed products.