Collaborative Filtering Models with the Credibility of Ratings

This paper quantifies the degree of user credit and adjusts it by specific items,and then gains accurate quantification on the credibility of ratings.In the case,the paper establishes the credibility matrix of ratings,and proposes four optimized similarity models based on credibility. The comparative experimental results, tested on different data sets,demonstrate that the newly proposed models have better performance in prediction accuracy of items at a reasonable cost in time,compared to traditional models.