生活消费平台虚假评论识别模型的研究  

Research of Fraud Review Detection Model on O2O Platform

作　　者：李晶[1] 吴国仕[1] 谢菲 姚旭[1] 齐佳音[3] 孙鹏飞[1] 

机构地区：[1]北京邮电大学软件学院,北京100876 [2]新华社通信技术局,北京100803 [3]北京邮电大学经济管理学院,北京100876

出　　处：《电子学报》2016年第12期2855-2860,共6页Acta Electronica Sinica

基　　金：国家973重点基础研究发展计划(No.2013CB329604);国家自然科学基金(No.71231002);北京市自然科学基金(No.9122018);教育部博士点基金(No.20120005110015);新华社713实验室技术研究项目--大数据与智能信息处理课题

摘　　要：生活消费平台已成为人们获取商家信息、反馈服务或产品质量的重要平台.虚假评论作为一种夸大或诽谤目标商家口碑的商业行为在生活消费平台很普遍,具有很强的危害性.本文对某网站的真实评论展开虚假评论研究,深入分析研究虚假评论的特征,从"可信度"的角度出发,提出用户及商家可信度模型.利用评论人的行为特征、商家的特征和评论文本的特征构建了虚假评论识别模型,经测试该模型达到了一个良好的识别效果.Living-consumption platform has become a very important platform for customers to extract information of businesses, and view or submit comments on the quality of services or products. It is common that fake reviews, as a com- mercial activity, are used to exaggerate or damage the reputation of a target business, which is extremely harmful. This paper chose an 020 （Online To Offline） platform,from which reviews are derived,to study fake reviews. With an in-depth study on features of fake reviews, it raised the user-credibility and shop-credibility evaluation model respectively from the credibili- ty perspective. Based on features of reviewers, businesses, and review texts, it established a fake review identification model, and through testing this model showed excellent performance in identification.

关 键 词：机器学习 虚假评论识别 可信度模型 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]