基于最小风险贝叶斯决策理论的在线评价排名方法研究  

Research on Online Evaluation Ranking Methods Based on Minimum Risk Bayesian Decision Theory

作　　者：田博[1] 陈舜杰 周雯 TIAN Bo;CHEN Shunjie;ZHOU Wen(School of Information Management&Engineering,Shanghai University of Finance and Economics,Shanghai 200433,China)

机构地区：[1]上海财经大学信息管理与工程学院,上海200433

出　　处：《上海管理科学》2018年第2期91-95,共5页Shanghai Management Science

基　　金：国家自然科学基金项目"BtoC电子商务中基于分层Bayesian网络的信任与声誉计算理论研究(编号:71302080)

摘　　要：随着电商模式不断发展和完善,在电商平台上出售的商品种类持续增加。消费者在做出购买决策之前,通常会参考其他消费者发表的在线评论。目前,电商客户评论中的好评与差评的权重风险是相等的,当顾客进行产品购买时会存在评论误判风险。针对好评与差评的权重风险相等的不足,本文提出了一种基于最小风险贝叶斯决策的评价排名方法。提出的方法首先利用网络爬虫语言抓取所有商品累计评论,形成TXT文档;然后根据已有的停用词词典,对生成的TXT文档进行文本预处理,并进行关键词的提取,通过关键词的分类将每一条评论生成一个文档,获得关键词权重;最后利用最小风险贝叶斯决策模型获得风险权重排名,并以天猫为例,与已有排名数据进行比较。提出方法新获得的排名为最小风险权重下的商家排名,有助于顾客在进行高风险商品购买决策时根据不同风险偏好,获得最优购买决策。With C2 Cmodel constantly development and completion,the diversity of goods is increasing as well.Most consumers will search others online comments before making their decisions,so the online comments which is one of the decision factors will have a huge impact on risk of consumer.Electricity suppliers sellers rank the positive comments and negative comments at the same possibility risk loss weight.However,there exists the misstatement risk when customers making decisions,and will have the different level of risk.This paper proposed a new method of the rank of Tmall shops based on minimum risk Bayesian decision theory to minimize the risk of misstatement.The outline of the proposed method is follows.Firstly,the online comments on e-commerce platform are grabbed as the source of data,and the comments with the method of JIEBA are pre-processed on python.Then base on minimum risk Bayesian decision rules,the weighted rank method with the minimum risk is proposed for the Tmall shops.At last,the computation results on concrete comments are given and compared with primary rank method used by Tmall.The new rank method based on minimum risk Bayesian decision theory will help the customer to make decision to low risk when buying high risk products,and optimize the decision at the same risk level.

关 键 词：最小风险贝叶斯决策 文本挖掘 在线评论 排名 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]