Collaborative Topic Regression-Based Recommendation Systems: A Comparative Study

摘要

The collaborative filtering is a very popular and powerful approach prominently being used in many research areas of computer science like recommendation systems, information retrieval, data mining, etc. When used for making recommendations, the traditional collaborative filtering methods suffer from certain problems, where the data sparsity is the significant one that causes the deterioration of the recommendation quality. In order to alleviate this issue, the research fraternity has started proposing the use of some additional domain information in formulating recommendations. In literature, different models have been proposed that make use of such kind of add-on information extensively and have also shown the promising performance than the other state-of-the-art approaches. Hence, the increasing use of add-on information is creating an overwhelming impact on the recommendation field. The piety of this article is to present a meticulous comparative study of various such recommendation models especially those which belong to the family of collaborative topic regression recommendation models in the light of several parameters and this study further leads to propose a novel recommendation prototype based on the fusion of different kinds of auxiliary domain knowledge.