基于协同过滤的电子商务个性化推荐算法研究  

Research on e-commerce personalized recommendation algorithm based on collaborative filtering

作　　者：成保梅[1] CHENG Baomei(Department of E?commerce,Business College of Shanxi University,Taiyuan 030031,China)

机构地区：[1]山西大学商务学院电子商务系，山西太原030031

出　　处：《现代电子技术》2019年第20期37-39,44,共4页Modern Electronics Technique

基　　金：山西大学商务学院2018年度院科研项目(2018016)

摘　　要：采用基于使用者的协同过滤推荐算法进行电子商务个性化推荐,将获取的评价数据作为推荐算法的输入,根据使用者行为的相似性获取“最近邻居”集,统计其中各邻居对项目商品的评价分数,并以使用者对项目商品的评分形式和使用者关注度最高的多个项目商品推荐列表形式进行项目商品推荐。在获取“最近邻居”集的过程中,通过使用者特征集提升电子商务推荐系统推荐最近邻居的准确度,利用兴趣度随时间变化函数修正使用者评价矩阵,从使用者特性和兴趣两方面对协同过滤推荐算法进行个性化改进。研究结果表明,所提算法推荐项目商品所需时间始终低于对比算法,且采用该推荐算法后电子商务平台交易成功率由38.4%提升至87.2%。The user?based collaborative filtering recommendation algorithm is used to carry out the e?commerce personal?ized recommendation,and the obtained evaluation data is used as the input of the recommendation algorithm.The"nearest neighbor"set is obtained according to the similarity of the users′behavior,and all the neighbors′evaluation scores for the proj?ect products are counted.The project product recommendation is carried out according to the users′scoring form on the project products and the recommendation list form of multi?project produts with the highest attention degree of users.In the process of obtaining the"nearest neighbor"set,the accuracy of recommending nearest neighbors of e?commerce recommendation system is improved by using users′feature sets,and user evaluation matrix is modified by means of the time?varying function of interest?ingness.The personalized improvement of collaborative filtering recommendation algorithm is conducted in the two terms of user characteristics and interests.The research results show that the proposed algorithm always takes less time to recommend project product than that of contrast algorithm,and the transaction success rate of e?commerce platform is increased from 38.4%to 87.2%.

关 键 词：电子商务 个性化推荐 协同过滤 商品推荐 个性化改进 交易平台 

分 类 号：TN99?34[电子电信—信号与信息处理] TP301[电子电信—信息与通信工程]