基于B2C帐号在线评论特征的聚类分析——以京东商城为例  

Clustering Analysis of Online Commenting Features Based on B2C Accounts——Taking JD. com as an Example

作　　者：段轶轩[1] 罗泽举[1] 

机构地区：[1]重庆工商大学数学与统计学院,重庆400067

出　　处：《重庆工商大学学报（自然科学版）》2014年第6期35-40,共6页Journal of Chongqing Technology and Business University:Natural Science Edition

基　　金：国家"十一五"科技支撑计划重大项目(2006BAJ05A06);电子商务及供应链系统重庆市重点实验室专项基金项目(2012ECSC0213);重庆工商大学创新型项目(yjscxx2013-025-09;yjscxx2012-037-036;yjscxx2012-037-037)

摘　　要：基于B2C帐号在线评论数据对帐号进行聚类,聚类算法采用EM、SimpleKMeans,并使用基于似然度的聚类算法评估准则作为聚类算法簇个数选取和算法比较依据。最终,选取EM算法并得到5类人群。得结论:站在帐号满意度的层面,京东商城的运营状况基本良好,但是,也有大概十分之一的高级账户存在流失的潜在风险;京东商城正在培育出一批忠诚度高、消费能力旺盛的客户群;评论字数较长的负面评论更易引起关注。Based on online commenting contents of B2C accounts to make clustering analysis of the accounts, by taking EM and SimpleKMeans as clustering algorithm,by taking likelihood based clustering algorithm estimation criterion as the basis for selecting the number of clustering algorithm clusters and algorithm comparison,five types of people group are generalized by selecting EM algorithm,and the conclusion is that the operating situation of JD. com is good but there is a potential risk for approximate 10 percent high-ranking customer accounts to run off at the level of accounts satisfaction degree,that JD. com is cultivating a group of customers with high-level loyalty and exuberant consumption capacity and that the negative comments with large number of words can attract more concerns.

关 键 词：B2C帐号 商品在线评论特征 EM算法 似然度的聚类评估准则 

分 类 号：G350.7[文化科学—情报学]