Research On Collaborative Filtering Recommendation Algorithm In E-Commerce System

With the popularization of Internet and the development of E-commerce, the method that focuses on SCD ( Web site centered design ) will be replaced with that focuses on UCD ( user centered design ). Presently recommender systems have gradually become an important part in E-Commerce IT technologies. As the most successful technology that is implemented in recommender system today, collaborative filtering has been attracted more and more attention by researchers.In this dissertation, based on the survey of collaborative filtering recommendation algorithms, some key challenges are analyzed posed in recommender systems. One of these challenges is ability of recommender systems to be adaptive to environment where users have many completely different interests or items have completely different content. Unfortunately, the traditional collaborative filtering recommendation algorithm can not make accurate recommendation for the cases. To address this issue, a novel collaborative filtering method has been explored, collaborative filtering based on user multiple interests. Finally, the results are experimentally evaluated, which suggest that approach is fairly effective. The contributions of this dissertation are as follows:First of all, by reviewing some of research literature related to collaborative filtering and recommender systems, taxonomy of collaborative filtering recommendation algorithms is extended, the existing algorithms of memory-based methods, model-based methods and hybrid methods are analyzed and compared. Therefore, a real-life problem is pointed out. That is user multiple-interests.Secondly, a novel algorithm is put forward based on user multiple-interests. The main idea is as follows: Firstly, sort the items into different categories for multiple-interests, then introduce user interests to differentiate and quantificationally measure predilection for item categories. Subsequently, built user clustering based on user interests and item categories. Finally, accomplish prediction and recommendation for user in different clusterings.Thirdly, the experiment conducted by applying MovieLens dataset indicated that new algorithm improved the sparsity of data and was more accurate than traditional methods.