关键词: 协同过滤, 有影响力近邻, 评分数量, 评分质量, 数据稀疏性

Abstract:

The recommendation performance of collaborative filtering is restricted by data sparsity. To solve this problem, the factor of user influence was thereafter defined according to the number of ratings to measure the relationship while calculating the similarity between users. Then, the influential user group was introduced according to the rating quality. On this basis, the chosen influential neighbor can work on the process of recommendations via combining the number of user ratings with the rating quality. Experiments show that the proposed method can significantly improve the recommendation performance.

Key words: collaborative filtering, influential neighbor, number of ratings, rating quality, data sparsity