微博垃圾账号检测研究

发布时间：2018-06-25 08:02

本文选题：异质信息网络 + 垃圾账号　；参考：《北京邮电大学》2017年硕士论文

【摘要】：近年来,随着诸如推特、新浪微博一类的新型社交网络的蓬勃发展,人们的生活和娱乐方式也发生了巨大的变化。社交网络为用户提供了在第一时间获取或发布多类型的信息如文本、图片、音频和视频的平台,在人们的日常生活中占有越来越重要的地位。然而,在社交网络为用户提供了信息交换平台的同时,垃圾账号的泛滥严重破坏了社交网络的生态平衡和用户体验。经过调研本课题将垃圾账号定义为主要以“僵尸粉”形式存在的、受机器控制的僵尸账号和以发送垃圾广告或其他垃圾信息为目的垃圾营销账号。本课题以新浪微博中的垃圾账号为研究对象,在对微博垃圾账号的行为特点的分析基础上,采用了一系列的垃圾账号检测方法,并最终通过异质信息网络实现了多类型数据结合的垃圾账号检测。本课题完成的工作分为以下几个方面:1.研究新浪微博数据的高效爬取和存储。2.对当前新浪微博平台中的垃圾账号的行为模式进行了分析和总结,并在此基础上进行了统计特征分析,使用从用户信息和博文信息中抽取的统计特征实现垃圾账号的检测。3.根据微博文本的特点如长度短、用词不规律以及特殊符号的使用等,设计了微博文本预处理过程和以词向量为基础的文本表示模型,通过特征选择和分类算法,实现基于文本的垃圾账号检测。4.研究异质信息网络和相关的相似度算法,构建了微博中的异质信息网络,提出了微博用户信息、博文信息和用户社交信息三类信息的合理结合方案以提升垃圾账号检测的效果,实验结果验证了该方法的有效性。
[Abstract]:In recent years, with the rapid development of new social networks such as Twitter and Sina Weibo, the way people live and play has changed dramatically. Social network provides a platform for users to obtain or publish many kinds of information such as text, picture, audio and video in the first time, which plays an increasingly important role in people's daily life. However, while social networks provide users with a platform for information exchange, the proliferation of spam accounts seriously undermines the ecological balance and user experience of social networks. After the investigation, the spam account is defined as the zombie account which exists mainly in the form of "zombie powder" and the spam marketing account for sending spam advertisement or other spam information. This subject takes the spam account in Sina Weibo as the research object, based on the analysis of the behavior characteristics of the Weibo spam account, a series of spam account detection methods are adopted. And finally through heterogeneous information network to achieve multiple types of data combined with spam account detection. The work accomplished in this project is divided into the following aspects: 1. Research Sina Weibo data efficient crawling and storage. 2. This paper analyzes and summarizes the behavior pattern of spam account in the current Sina Weibo platform, and analyzes the statistical characteristics on the basis of which, using the statistical features extracted from user information and blog post information to realize the detection of spam account. 3. According to the characteristics of Weibo texts, such as short length, irregular use of words and the use of special symbols, the preprocessing process of Weibo text and the text representation model based on word vector are designed, and the algorithm of feature selection and classification is adopted. Implementation of text-based spam account detection. 4. This paper studies heterogeneous information network and related similarity algorithms, constructs heterogeneous information network in Weibo, and proposes a reasonable combination of Weibo user information, blog information and user social information to improve the effect of spam account detection. The experimental results show that the method is effective.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.092

【相似文献】