中文摘要：

电子商务水军的识别已成为众多网络水军识别领域中较为突出的研究问题.已有电子商务水军识别研究多关注网络水军自身属性和行为特征,以发现隐藏其中的网络水军行为模式.电子商务服务急速扩张刺激以获取经济利益为目的的大规模水军傀儡账号泛滥,水军逐渐形成团体规模.大规模水军泛滥使得单独网络水军的行为更加趋向正常用户,团体内部成员间不具有明显相似性,基于特征模式识别的研究方法无法很好地发现该类电子商务水军.文中定义电子商务用户的加权用户关系图模型,分析其谱特征定位用户关系图中的异常关系结构,从而找出隐藏其后的大规模电商水军团体,并提出一种基于用户关系图模型定位大规模电商水军团体的算法.文中在两个国内外最具代表性的电子商务平台（淘宝、亚马逊）数据集上进行了大量实验,并评估了算法的不同参数定位电商水军团体的能力.实验结果表明文中提出的加权用户关系图异常结构能够很好地定位隐藏较深的大规模电子商务网络水军团体,且加权用户关系图定位电子商务网络水军团体的能力优于非加权用户关系图.

英文摘要：

With rising popularity,the identification of online shopping websites spammers has become one of the hottest topics in the wide field of spammer detection.The existing works in detection of online shopping websites spammers usually focus on the behavior patterns of spammers.When most spammers start functioning as many large spammer groups to game the detection system,the individual member of the large group tends to act like a normal user.The behavior pattern discovery of spammers cannot effectively detect these large spammer groups.This paper proposes a novel angle to this problem by modeling the weight relation network of online shopping websites users.An algorithm called User Relation Graph Spectrum-based Spammer Group Detection（URGSSGD）is also proposed to detect the signal network of large spammer group in the user relation network by analyzing its spectrum features.Experiments on two typical real-life review datasets, which comprise Amazon and Taobao review dataset, demonstrate the effectiveness of proposed model which outperforms the existing methods to identify large spammer groups.Also this paper proves the weight user relation network performs well than the unweight user relation network on detecting large spammer groups in online shopping websites.