用户聚类的电子商务推荐系统研究

Research on User Clustering of Recommendation System in E-commerce

随着电子商务规模越来越大,协同过滤推荐算法的可扩展性差的问题也越来越受到人们的重视,提出了一种基于用户项目类偏好值矩阵聚类的合作推荐方法,解决了"冷开始"问题,并且由于只在目标用户所属类别中搜索其最近邻居,减少了搜索空间,有效地提高了推荐系统的实时响应速度.

Nowadays, with the scale of E-commerce is getting larger and larger, more importance has been attached to the problem of poor expansibility appearing in collaborative filtering recommendation al- gorithm. This paper describes a method of cooperative recommendation based on user-item preference values and matrix clustering, which has solved ＂cold start＂ problem to some extent. This method only searching for its nearest neighbor in the classification of object user, so the search space is reduced and the real-time performance of the recommendation system can be effectively improved.