Research On Collaborative Filtering Recommendation Algorithm Based On Clustering And User Interest

In the end of twentieth century, the personalized recommendation technology was putforward as an independent concept. Then, a variety of recommendation techniques have developedrapidly in recent years. For example, information retrieval, content-based recommendation,association rules-based recommendation, collaborative filtering technology and mixed recom-mendation. collaborative filtering algorithm can recommend cross-commodity type products tousers and can help users to find new information that they are interested in. The finalrecommences are based on the basis data of User-Item rate matrix which can give full play to theinitiative of users. This technology is the best one of personality recommendation algorithm. Withthe number of users and commodities of the e-commerce platform increasing sharply, there aresome problems of collaborative filtering such as cold start, the sparse of User-Item rate matrix, thescalability of system to be solved.To solve the existing problems of collaborative filtering, the paper presents a series ofcorresponding solutions. After research of the paper, we find that clustering technology can beused to solve the problem of the scalability of the system, we use an improved k-means algorithmwhich can solve the problem of the scalability of the system and can reduce the number of isolatedpoints; the traditional similarity measurement method is difficult to accurately reflect thesimilarities among the users and it can’t reflect the user interest changes in time. A novel similaritymeasurement algorithm is put forward in the paper which can ameliorate the two drawbacks inexisting algorithms and improves the recommendation accuracy.Experiment results show that the novel algorithm ameliorates the drawbacks in existingalgorithms and improves the recommendation accuracy. The feasibility and effectiveness of theimproved algorithm is certified.