Research On Personalized Recommender System Based On Collaborative Filtering Algorithm

Information overload and information loss have become the bottlenecks for people to make effective use of the information on internet. Information overload is one of the negative effects of the information extreme abundance in the information age. As the application of internet rapidly develops and the net information increases quickly, the users find it hard to timely digest and absorb the too much information. Information loss means that users do not know how to definitely express the need of the internet information or how to correctly and effectively find the information they are interested due to the universality of the information resources on internet. Now, most search engines are lack of initiative. With no consideration for the users'interest and hobby, they cannot effectively solve the problems of information overload and information loss.Recommender system is the technology derived on the base of information filtering technology. The information filtering technology can solve relatively better the problems of "information overload" and "information loss",_so that people can fully use the internet information. The information filtering technology is the base of realizing information resources personalized recommendation. However, as the internet information scale constantly expands, the recommender system faces a series of problems as well. This thesis will have a research on the personalized recommender system algorithm by combining the existing internet resources.The thesis firstly introduces WEB mining and personalized recommender system. Then, it introduces the theory and application of collaborative filtering algorithm, and analyzed, points out the problems existing in this algorithm. It provides the collaborative filtering algorithm basing on WEB diary and cluster analysis. The algorithm will analyze the server WEB diary offline and collect the users with similar interests in groups based on the method of user cluster. By finding neighbors with the similar interests, when online, the system will only recommend the target users the commodities they are likely to show interest so as to improve the online data processing efficiency. Finally, the algorithm is applied in the WEB_site which the thesis studied, so as to prove that the algorithm can effectively improve recommend quality and recommend efficiency of the recommender system.