Improvement of similarity algorithm in Collaborative Filtering based on Stability Degree

With the fast development of World Wide Web, Web-based applications and services should allow user to get the right personalized information quickly and effectively. Collaborative Filtering acts a very important role in web service personalization and Recommender System. In this paper, Stability Degree was proposed to improve the accuracy of collaboration filtering both based on item and user, kinds of Stability Degree were introduced into item and user similarity computation, and the results show that the prediction accuracy can be improved from 10 percents to 25 percents in different case, using this improved similarity algorithm, MAE can be also reduced faster than classic methods.