点评类社区评论效用分类模型研究

A Study on Review Utility Classification Model of Online Review Community

通过数据挖掘的方法对点评类社区的评论进行分类,挖掘效用高的评论以此降低用户信息搜寻成本。构建在线评论效用指标体系,利用k-means聚类将评论分为4类,利用关联规则挖掘类内特征,构成在线评论二维分类模型并通过大众点评网的数据进行实验。将在线评论分为高效评论、可疑评论、潜在评论以及无效评论4类,重点挖掘可疑评论以及潜在评论的特征并提出具有针对性的措施。由于匿名评论者的信息无法获取,因此未将匿名评论考虑在内。该模型是一种简单有效的在线评论效用分类模型。

Reviews of online products have caused information overload problem,which makes that we must find high utility reviews to help users in review community to cut search costs.This paper designs an index system for review utility and takes advantage of data mining algorithms such as k-means clustering and association rule to classify them.Moreover,the data collected from dianping.com is used to test this classification model.Reviews are divided into four categories,which are accordingly named efficient comment,suspicious comment,potential comment and invalid comment.Particularly,we＇ve analyzed the suspicious comment and potential comment by association rule and brought forward countermeasures accordingly.Anonymous comments haven＇t been taken into account.We provide a simple and efficient way to classify reviews of online products.