Collaborative Filtering Applied Research In E-commerce Recommendation System

Recently, personalized recommender systems, especially collaborative filtering recommender systems, have achieved widespread successes on the Web. The tremendous growth in the amount of available information and the kinds of commodities to Web sites poses some key challenges for recommender systems, so the problems of cold-start and sparsity in collaborative filtering recommendation are needed to be solved.To address these issues, the personalized recommender systems and some recommendation methods are introduced, in this paper, and the problems in collaborative filtering recommendation are analyzed. To efficiently resolve these problem that the new items are difficult to recommend in collaborative filtering algorithm. A new method based item matrix partition is proposed. The essential idea was that the item matrix can be partitioned by using classification tree algorithm and get low-dimensional matrices. This method predicts new item rating based on item rating that users have rated and item similarity. Compared traditional collaborative filtering method, the experimental results show that our approach can find a solution to the problem of new item recommendation effectively.In this foundation, This study applies "Collaborative Filtering", a kind of web-personalization, to recommend buyers the products according to their similar group and provide sellers the potential customers. Moreover, for the recommandation freshness, this study also recommends the correlative up-to-date products by content-based approach.After the empirical test in an e-Marketplace of food industry, this study finds that "Collaborative Filtering Matching System" can make a goodmatching performance and add recommend efficiency for E-market , sellers can benefit from this system.