Improvement of Collaborative Filtering with the Simple Bayesian Classifier

Collaborative-filtering-enabled Web sites that recommend books, CDs, movies, and so on, have become very popular on the Internet. Such sites recommend items to a user on the basis of the opinions of other users with similar tastes. In this paper, we discuss an approach to collaborative filtering based on the Simple Bayesian Classifier, and apply our model to two variants of the collaborative filtering. One is user-based collaborative filtering, which makes predictions based on the users’ similarities. The other is item-based collaborative filtering, which makes predictions based on the items’ similarities. In our approach, the similarity between users or items is calculated from negative ratings and positive ratings separately. To evaluate our algorithms, we used a database of movie recommendations. Our empirical results show that our proposed Bayesian approaches outperform typical correlation-based collaborative filtering algorithms. We also discuss an approach that combines user-based and item-based collaborative filtering with the Simple Bayesian Classifier to improve the performance of the predictions. After the user-item rating matrix has been filled out with pseudo-scores generated by the item-based filter, the user-based recommendation is applied to the matrix. We show that the combined method performs better than the single collaborative recommendation method.