Author(s)

Zeliang Hao

Corresponding Author

Zeliang Hao

ABSTRACT

Personalized recommendation system is a very common system at present, which partly solves the problem of information overload. Among them, collaborative filtering technology is the most widely used. However, collaborative filtering technology always has some problems, such as cold start, sparse data and so on. In order to solve these problems, we have to use some method to improve the accuracy of system prediction, which involves the calculation of similarity, fuzzy clustering, user profile, item characteristics and other problems. This paper briefly introduces the relevant background knowledge of collaborative filtering technology, summarizes some similarity algorithms, summarizes and analyzes some methods to overcome data sparsity, and discusses two of them: weighted joint similarity algorithm based on user interest and Pearson_ After _ SVD algorithm. Then, some experimental results of the two methods are compared, and a new method is proposed: the similarity algorithm of the first method is fused to the similarity algorithm of the latter one. In the future work, it is necessary to constantly put forward new methods to give users a better recommendation experience.

KEYWORDS

Collaborative filtering, sparse data, Pear_ After_ SVD, Weighted Similarity