Improved Collaborative Filtering Algorithm based on a Novel Similarity Measure for Recommender Systems
Yannian Chen, Fei Chu, Mingyuan Wang, Linjia Sun
Faculty of Computer Science and Technology, Anhui University, Hefei, 230601, China
Abstract: With the explosive growth of services and items on the internet, recommender systems have been widely used to recommend personalized items to users. Collaborative filtering (CF) is one of the most suc-cessful recommendation techniques used in recommender systems. Similarity computation is the critical step, which will significantly affect the predictive accuracy of CF. Traditional similarity measures, such as Pear-son’s correlation coefficient (PCC) and cosine similarity (COS), have inherent limitations on accuracy. And there are still many problems in the improved measures proposed recently in some papers. To address these problems, we fully consider different situations that may be encountered in the process of recommendation and propose a novel similarity measure. Based on the proposed similarity measure, we propose a new colla-borative filtering algorithm to improve the accuracy of recommendation and meanwhile enhance the robust-ness of recommender systems. The results of experiments conducted on three real datasets prove that our CF algorithm achieves excellent performances.
Keywords: Recommender system; Similarity measure; Collaborative filtering; Predictive accuracy; RDCF