Effective Matrix Factorization for Online Rating Prediction

Recommender systems have been widely utilized by online merchants and online advertisers to promote their products in order to improve profits. By evaluating customer interests based on their purchase history and relating it to commodities for sale these retailers could excavate out products which are most likely to be chosen by a specific customer. In this case, online ratings given by customers are of great interest as they could reflect different levels of customers' interest on different products. Collaborative Filtering (CF) approach is chosen by a large amount of web-based retailers for their recommender systems because CF operates on interactions between customers and products. In this paper, a major approach of CF, Matrix Factorization, is modified to give more accurate recommendations by predicting online ratings.