Recommendation by Mining Multiple User Behaviors with Group Sparsity


  • Ting Yuan Chinese Academy of Science
  • Jian Cheng Chinese Academy of Science
  • Xi Zhang Chinese Academy of Science
  • Shuang Qiu Chinese Academy of Science
  • Hanqing Lu Chinese Academy of Science



recommender systems, Matrix Factorization


Recently, some recommendation methods try to improvethe prediction results by integrating informationfrom user’s multiple types of behaviors. How to modelthe dependence and independence between differentbehaviors is critical for them. In this paper, we proposea novel recommendation model, the Group-Sparse MatrixFactorization (GSMF), which factorizes the ratingmatrices for multiple behaviors into the user and itemlatent factor space with group sparsity regularization.It can (1) select out the different subsets of latent factorsfor different behaviors, addressing that users’ decisionson different behaviors are determined by differentsets of factors;(2) model the dependence and independencebetween behaviors by learning the sharedand private factors for multiple behaviors automatically; (3) allow the shared factors between different behaviorsto be different, instead of all the behaviors sharingthe same set of factors. Experiments on the real-world dataset demonstrate that our model can integrate users’multiple types of behaviors into recommendation better,compared with other state-of-the-arts.




How to Cite

Yuan, T., Cheng, J., Zhang, X., Qiu, S., & Lu, H. (2014). Recommendation by Mining Multiple User Behaviors with Group Sparsity. Proceedings of the AAAI Conference on Artificial Intelligence, 28(1).