Research Of Collaborative Filtering Recommendation Based On Tag Clustering And Item Topic

With the rapid development of information technology,people must face the increasingly serious problems of information overload while enjoying the convenience of the Internet.It is difficult for users to find interested information from the massive network resources quickly and exactly.In this context,personalized recommendation system is proposed as an effective method to solve the problems above.Collaborative filtering recommendation is one of the most widely used algorithms among all recommended algorithms,but it still has data sparse and cold start problems.As an important product in web 2.0,social tagging system provides tags as highly valuable data to recommender system.Tags can be used as supplement information to obtain the preference of users and characteristics of resources accurately.The combination of personalized recommendation algorithm with social tags can effectively improve the efficiency of personalized recommendation.This paper analyzed the principle of item-based collaborative filtering algorithm,used tag clusters to represent item topics,proposed the collaborative filtering recommendation algorithm based on tag clustering and item topic aiming at solving the problems about data sparse and ignoring the characteristics of items.Experiment results show the algorithm is valid.This paper also designed a collaborative filtering recommendation system prototype based on the proposed algorithm.The main work of this paper is as follows.(1)Elaborating the researches and achievements of the recommendation system by domestic and foreign scholars.Analyzing the principle and related technologies of recommendation system and emphasizing the mainstream recommendation algorithms and its limitations.(2)After the introduction of the social tagging system,this paper analyzed the research status and problems,such as massive tags and semantic fuzziness of tag-based personalized recommendation,used the algorithm which is proposed for finding communities in complex network to cluster social tags in order to eliminate the semantic fuzziness.(3)Aiming at solving the inaccuracy of neighbor selection problems in traditional item-based collaborative filtering algorithm caused by data sparse and algorithm strategy itself,this paper proposed an improved collaborative filtering algorithm basedon tag clustering and item topic,regarded the tag clusters as different topics,combined the rating data with item topics to calculate the similarity between items.The effectiveness of this algorithm is verified compared with other algorithms.(4)This paper designed a collaborative filtering recommendation system prototype based on software engineering theory and the proposed algorithm above,described the general design of the system and detailed design of main modules.