电子商务推荐系统中推荐策略的自适应性  

Self-adaptive recommendation policy in recommender systems

作　　者：刘平峰[1] 聂规划[1] 陈冬林[1] 

机构地区：[1]武汉理工大学经济学院,武汉430070

出　　处：《计算机工程与应用》2007年第4期23-25,共3页Computer Engineering and Applications

基　　金：国家自然科学基金(the National Natural Science Foundation of China under Grant No.70572079)。

摘　　要：针对电子商务推荐系统中各种推荐技术的不足,提出推荐策略的自适应方法。用二元组<用户知识,推荐商品>代表推荐环境的根本特征,采用ART神经网络进行自学习,获取推荐环境的不同聚类,每个聚类代表了某种推荐环境,对推荐结果的反馈情况进行统计分析,确定每个聚类的最佳推荐技术。向用户推荐商品时,根据用户所在聚类采用具有最佳推荐质量的推荐技术向用户作出推荐。整个系统的工作过程不需要人工干预,具有自适应性。A method of choosing recommendation policy adaptive to the recommendation context is proposed in this paper to overcome the shortcomings of different recommendation techniques.The recommendation context is denoted as a duple 〈user knowledge,commodity category〉.The clusters of the recommendation context are acquired by exploiting the ART artificial neural network.The recommendation technique of best recommendation quality corresponding to each cluster is acquired by the analysis of reflection from users on the recommendation results.So the recommendation technique of best quality can be applied to generate recommendation results for user in correspondence to the cluster in which the duple 〈user knowledge,commodity category〉 of the user reside in.The working process of the system is self-adaptive without manual intervention.

关 键 词：电子商务 推荐系统 推荐策略 自适应 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]