一种新的基于非对称皮尔逊相似度的协同过滤算法

In traditional collaborative filtering algorithms, the similarity between users is usually calculated by Cosine similarity, Pearson similarity, JacPearson similarity, etc. However, for highly sparse user data matrix, the traditional similarity algorithm has some shortcomings, which will cause the users have a much higher similarity. In this paper, through qualitative and quantitative analysis of traditional similarity algorithms, their advantages and disadvantages were analyzed before putting forward a new asymmetric Pearson similarity algorithm (ASC-Pearson). Experiment with MovieLens dataset shows that, for condition with sparse user data, the new similarity algorithm performs better than the traditional similarity algorithms.