SVMs Classification Based Two-side Cross Domain Collaborative Filtering by inferring intrinsic user and item features

作者:

Highlights:

摘要

Recently, Cross Domain Collaborate Filtering (CDCF) is a new way to alleviate the sparsity problem in the recommender systems. CDCF solves the sparsity problem by transferring rating knowledge from auxiliary domains. Most of previous work only uses one-side (user-side or item-side) auxiliary domain information to help the recommendation in the target domain. In this paper, we propose a two-side Cross Domain Collaborate Filtering model. We assume that there exist two auxiliary domains, i.e., user-side domain and item-side domain, where the user-side auxiliary domain shares the same aligned users with the target domain, and the item-side shares the same aligned items. Also both the two auxiliary domains contain dense rating data. In this scenario, we first employ the bi-orthogonal tri-factorization model to infer the intrinsic user and item features from the user-side and item-side auxiliary domain respectively. The inferred intrinsic features are independent on domains. Then we convert the recommendation problem into a classification problem. In detail, we use the inferred user and item features to compose the feature vector, and use the corresponding rating as the class label. Thus the user-item interactions can be represented as training samples. Finally, we employ SVMs model to solve the converted classification problem. The major advantage of our model is that it can make use of both user-side and item-side shared information. Furthermore, it can infer the domain independent user and item features. Thus it can transfer knowledge from auxiliary domains more effectively. We conduct extensive experiments to show that the proposed model performs significantly better than many state-of-the-art single domain and cross domain CF methods.

论文关键词:Cross domain collaborative filtering,Transfer learning,Bi-orthogonal tri-factorization,Classification problem

论文评审过程:Received 24 May 2017, Revised 7 November 2017, Accepted 10 November 2017, Available online 11 November 2017, Version of Record 19 December 2017.

论文官网地址:https://doi.org/10.1016/j.knosys.2017.11.010