摘要
提出了一种新的混合型推荐系统框架,该框架采用改进的K-means方法对用户和产品进行聚类,然后用Logistic回归对交易数据进行关联分析,最后使用线性信息融合模型对所有规则进行综合判断,给出合理的推荐结果.介绍了这个框架的实现过程,并将系统应用于一个具体的商业案例,以对推荐效果进行检验.结果表明:新系统适用于垂直型电子商务网站,推荐准确度较高.
This paper presents a novel hybrid recommender system for e-commerce.The following three phases are approached,clustering the customers and the products with an improved K-means algorithm, analyzing the association rules on transaction history using the logistic regression,and integrating all rules in a linear information fusion model,to achieve a valuable recommendation result.The implementation on a practical business application of this system is also introduced in this paper.The performance evaluation validates the recommendation result,and proves that our system works better in the sort of vertical e-commerce websites with higher prediction accuracy.
出处
《系统工程理论与实践》
EI
CSSCI
CSCD
北大核心
2010年第5期928-935,共8页
Systems Engineering-Theory & Practice
基金
厦门大学国家211三期项目:立体通信与信息集成智能技术
关键词
推荐系统
电子商务
决策支持系统
信息融合
recommender system
e-commerce
decision support system
information fusion