基于对话流主题的异常社交群组发现研究
首发时间:2017-12-08
摘要:随着移动互联网的快速发展,各类在线社交聊天群组数量以及群组规模不断扩大,人们的社交生活得到了极大丰富。然而由于社交群组规模庞大、隐蔽性强,现有监督管理方式存在漏洞,容易被不法分子绕过,从而导致各种不安全事件频发,对社交群组管理也提出了更高的管理需求。为了解决上述问题,本文提出一种基于对话流和群描述主题的异常群组发现方法,通过对社交聊天群组的对话流和群描述信息进行采集,利用Latent Dirichlet Allocation(LDA)算法对群描述信息和对话流信息进行主题提取,然后通过比较同一社交群组的对话流和群描述主题来识别异常社交群组。本文对QQ应用下的300个QQ群组进行实验,共检测出66个异常社交群组,准确率较高,因此,本文提出的方法能够很好的发现异常社交群组。
For information in English, please click here
Research on Abnormal Social Groups Based on conversation flow topics
Abstract:With the rapid development of the mobile Internet, the number of online social chat groups as well as the size of them are expanding, and people\'s social life has been greatly enriched. However, due to the large size and concealment of social groups, there are loopholes in the supervision and management methods, thus they are easily circumvented by criminals, resulting in frequent occurrence of various unsafe incidents and higher management requirements for social group management. In order to solve the above problems, this paper proposes an abnormal group discovery method based on the conversation flow and group description topic. After collecting the conversation flowsand descriptions of social groups, this paper uses the Latent Dirichlet Allocation (LDA) algorithm to extract the topics of descriptions and conversation flows, and then identifies abnormal social groups by comparing the conversation flow and description topics of the same social group. This paper experiments on300 QQ groups under QQ application, and detectes 66 abnormal social groups with high accuracy. Therefore, the proposed method can find abnormal social groups well.
Keywords: Social Group Anomaly Detection LDA QQ Group
基金:
论文图表:
引用
No.****
同行评议
共计0人参与
勘误表
基于对话流主题的异常社交群组发现研究
评论
全部评论0/1000