PROJECT TITLE:

Relational Collaborative Topic Regression for Recommender Systems - 2015

ABSTRACT:

Because of its successful application in recommender systems, collaborative filtering (CF) has become a hot analysis topic in knowledge mining and information retrieval. In traditional CF ways, solely the feedback matrix, that contains either specific feedback (also called ratings) or implicit feedback on the things given by users, is employed for coaching and prediction. Usually, the feedback matrix is sparse, that means that almost all users interact with few things. Due to this sparsity drawback, traditional CF with solely feedback info will suffer from unsatisfactory performance. Recently, many researchers have proposed to utilize auxiliary info, such as item content (attributes), to alleviate the data sparsity problem in CF. Collaborative topic regression (CTR) is one of those ways which has achieved promising performance by successfully integrating both feedback data and item content information. In many real applications, besides the feedback and item content information, there may exist relations (conjointly referred to as networks) among the items which can be useful for recommendation. During this paper, we have a tendency to develop a completely unique hierarchical Bayesian model known as Relational Collaborative Topic Regression (RCTR), that extends CTR by seamlessly integrating the user-item feedback information, item content information, and network structure among items into the identical model. Experiments on real-world datasets show that our model will achieve higher prediction accuracy than the state-of-the-art ways with lower empirical training time. Moreover, RCTR will learn sensible interpretable latent structures which are useful for recommendation.