Abstract: To measure the similarity of the users is the core of collaborative filtering algorithm. The similarity of users has a significant impact on the results of personalized recommendation. In this paper, method of similarity measurement is optimized through analyzing users - item rating records and comparing with pearson and jaccard similarity. This optimization method is applied to empirical analysis of the data set provided by the MovieLens site. Empirical studies show that the new algorithm can improve the accuracy of personalized recommendation and overcome the impact of data sparsity on recommendation quality to some extent.

Key words: collaborative filtering, similarity, personalized recommendation

摘要： 衡量用户的相似性是协同过滤算法的核心内容，用户间相似性的准确率对个性化推荐的结果会有显著影响。通过对用户-项目评分记录的分析，在比较pearson和jaccard相似性的基础上对相似性度量方法进行改进，并将该改进方法应用于MovieLens站点提供的数据集进行实证分析。实证研究表明，改进后的算法可以提高个性化推荐的准确性，并在一定程度上克服数据稀疏性对推荐质量的影响。

关键词: 协同过滤, 相似性, 个性化推荐