Department of Computer Science
Heterogeneous data fusion via matrix factorization for augmenting item, group and friend recommendations
Up to now, more and more social media sites have started to allow their users to build the social relationships. Take the Last.fm for example (which is a popular music-sharing site), users can not only add each other as friends, but also join interest groups that include people with common tastes. Therefore, in this environment, users might be interested in not only receiving item recommendations, but also friend recommendations whom they might consider putting in the contact list, and group recommendations that they may consider joining in. To support such needs, in this paper, we propose a generalized framework that provides three different types of recommendation in a single system: recommending items, recommending groups and recommending friends. For each type of recommendation, we in depth investigated the algorithm impact of fusing other two information resources (e.g., user-item preferences and friendship to be fused for recommending groups), along with their combined effect. The experiment reveals the ideal fusion mechanism for this multi-output recommender, and validates the benefit of factorization model for fusing bipartite data (such as membership and user-item preferences) and the benefit of regularization model for fusing one mode data (such as friendship). Moreover, the positive effect of integrating similarity measure into the regularization model is identified via the experiment. Copyright 2013 ACM.
Friend recommendation, Group recommendation, Item recommendation, Matrix factorization, Regularization
Source Publication Title
Proceedings of the 28th Annual ACM Symposium on Applied Computing
Link to Publisher's Edition
Zeng, W., & Chen, L. (2013). Heterogeneous data fusion via matrix factorization for augmenting item, group and friend recommendations. Proceedings of the 28th Annual ACM Symposium on Applied Computing, 237-244. https://doi.org/10.1145/2480362.2480415