基于客户行为和兴趣变化的电子商务推荐系统  

An E-commerce recommender system based on customers' behaviors and interest drifts

作　　者：张红霞[1] 杨渊[2] 郎维[3] 

机构地区：[1]安徽理工大学经济与管理学院,安徽淮南232001 [2]淮南师范学院经济与管理学院,安徽淮南232001 [3]吉林大学中日联谊医院,吉林长春130033

出　　处：《宝鸡文理学院学报（自然科学版）》2012年第2期52-56,共5页Journal of Baoji University of Arts and Sciences(Natural Science Edition)

基　　金：安徽省教育厅人文社科研究项目(2011sk147)

摘　　要：目的改善传统协同过滤推荐系统,提高推荐结果的准确性。方法采用混合推荐方法,将客户对商品的显式评分和客户的访问行为相结合,并借鉴心理学遗忘规律,引入指数遗忘函数,将客户评分数据的重要性依据时间逐渐衰减,以此来反映客户的兴趣变化。结果设计了基于客户行为和兴趣变化的混合推荐模型,给出了客户兴趣数据的获取方法和考虑客户兴趣变化的推荐方法。结论该方法在推荐结果的准确性上要优于传统的协同过滤推荐。Aim To improve the traditional collaborative filtering（CF） recommender system and raise the accuracy of recommended results.Methods A hybrid recommender approach is proposed which combines the customers’ explicit ratings on the goods and the customers’ access behaviors on the website.Then,with reference to the forgetting law of psychology,an index forgetting function is introduced to diminish the importance of the ratings according to the time and to reflect the changes of customers’ interests.Results A hybrid recommender system model based on customers’ behaviors and interest drifts is designed.The methods to get the customers’ interest data and to carry out the recommendation considering the changes of customers’ interests are given.Conclusion The new approach is more accurate than the traditional CF method.

关 键 词：推荐系统 协同过滤 访问行为 兴趣变化 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]