Abstract: Collaborative filtering is the most successful technology for building recommendation systems. But with a large number of users and items, this method faces serious problems such as sparsity which makes the recommendation efficiency decline linearly. In this paper a concept hierarchy methodology ameliorating user-item matrix was suggested. By using buy-data and click-through-data and integrating items of similar users and those of multi-level association, this method showed good performance on sparsity set.

Key words: ecommendation system, concept hierarchy, data mining, collaborative filtering

摘要：  协同过滤算法(collaborativefiltering)目前较为成功地应用于个性化推荐系统中,但随着系统规模的扩大,面临很严重的稀疏性问题,制约了推荐效果。文中提出概念分层的方法对用户项矩阵进行改进,同时使用交易数据和点击流数据,将相似用户选择项与多层次关联规则推荐项相结合,在稀疏数据集上表现出较好的性能。

关键词: 推荐系统, 概念分层, 数据挖掘, 协同过滤