在线产品虚假评论检测技术研究

Study of Online Product Fake Review Detection

网上购物平台的商家通过虚假评论对自家商品进行美化,并对竞争商品进行抹黑。目前虚假评论人群组的欺诈特征比单个虚假评论人明显,因此,大量的虚假评论检测方法以对虚假评论人群组检测为主。本文提出一种新的在线产品虚假评论检测方法,该方法采用时间窗口提取时间序列上的评论数据,使时间窗口内的评论数据生成二部图结构;将二部图结构转化为评论人图结构,然后用SCAN算法对评论人图进行聚类,并计算聚类后评论人群组特征;最后使用支持向量机对已标注的评论人群组特征数据训练分类器,并对未标注的进行检测。实验结果表明,该方法可以通过时间窗口对正在发生的虚假评论欺诈活动进行有效检测。

Online merchants beautify their own products through review spam and smear competing products.At present,the feature of review spammers group is more obvious than that of single review spammers.Therefore,a large number of review spam detection methods are mainly used for review spammers group detection.This paper proposes a new online product review spam detection method,which adopts the time window to extract the reviews on the time series data,make reviews in time windows data generation figure structure,and then transform two parts graph structure into reviews people figure structure,and then use the SCAN algorithm for reviews graph clustering,and calculate the clustering characteristics after reviews groups,finally use the SVM classifier.The experimental results show that the proposed method can be used to detect the review spam activity of reviews online through the time window.