协同过滤推荐算法.doc

A Collaborative Filtering Recommendation Algorithm Based on Interest Forgetting Curve Abstract Collaborative filtering algorithm is one of the most successful technologies used in personalized recommendation system. However, traditional algorithms focus only on the user ratings and do not take the changes of user interest into account, which affect recommeandation quality seriously. Based on experiment, this paper first explored the law of user interest changing—interest forgetting curve. Then, use lately rated items to represent the user’ current interest; for each historily visited item, calculate the integrated data weight based on interest forgetting curve and the rating matrix; for each item without the user’ score, calculate prediction based on item similarity and item integrated data weights. While calculating items’ similarity, this paper compounded item attribute similarity and item score similarity, which was more comprehensive and accurate. The experimental results show that the proposed algorithm can provide better recommandation precision and recall ratio. Keywords： collaborative filtering (CF), similarity, Interest forgetting curve Introduction With the development of information technology, E-commerce has become an integral aspect of doing business. Because of the convenience of the internet, a tremendous amount of product-related information is available to customers at the very low cost, which make the users even hard to choose the products according to their references. This is called information overload. To address the issue, recommendation systems have been widely used at many large electronic commerce sites to suggest products, services to potential customers. For example, companies such as A, N, H, and CDNOW have successfully implemented commercial recommendation systems[1]. Two main technologies are usually adopted in personalized service systems: content-based filtering and collaboraive filtering(CF). Content-based filtering methods provide