基于三维协同过滤的C2C电子商务推荐系统  

C2C e-commerce recommender system based on three-dimensional collaborative filtering

作　　者：艾丹祥[1] 左晖[2] 杨君[1] 

机构地区：[1]广东工业大学管理学院,广东广州510520 [2]广东工业大学经济与贸易学院,广东广州510520

出　　处：《计算机工程与设计》2013年第2期702-706,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(70971027);国家社会科学基金项目(11CTQ020);广东省自然科学基金项目(10451009001004318)

摘　　要：个人对个人电子商务(customer to customer,C2C)是目前主流的电子商务模式之一,为解决C2C电子商务网站中特殊的推荐问题,对传统的二维协同过滤方法进行了扩展,提出了能进行卖家和商品组合推荐的三维协同过滤推荐方法,并在此基础上设计了C2C电子商务推荐系统,阐述了该系统的基本架构和推荐过程中的关键运算。该系统利用卖家属性计算卖家相似度,并依据销售关系和卖家相似度对评分数据集进行填充,以解决三维评分数据的稀疏问题;采用协同过滤思想,利用历史评分计算买家相似度,获取最近邻并预测未知评分,最终将预测评分最高的卖家和商品组合推荐给目标买家。实验结果表明,该系统具有较好的推荐效果。C2C （Consumer to Consumer） ecommerce is currently one of the main ecommerce patterns. To solve the special rec ommendation problem in C2C ecommerce websites, a threedimensional collaborative filtering recommendation approach which can recommend seller and product combinations is proposed by extending the traditional twodimensional collaborative filtering approach. And a C2C ecommerce recommender system based on the proposed approach is designed. The framework of the sys tem and the crucial calculations in the recommendation process are discussed. The system firstly calculates seller similarities using seller features, and fills the ratings set based on sales relations and seller similarities to solve the sparsity problem of the threedimensional rating data. Then it calculates the buyer similarities using historical ratings, decides neighbors and predicts un known ratings by applying collaborative filtering principle. Finally it recommends the seller and product combinations with the highest prediction ratings to the target buyer. The good recommendation performance of the system is also proved by a true data experiment.

关 键 词：个人对个人电子商务 电子商务推荐 三维推荐 协同过滤 推荐系统 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]