一种基于微博用户行为的僵尸粉识别方法  

A recognition method of zombie fans on micro-blog user's behavior

作　　者：张锡英[1] 车鑫[1] 田宪允 

机构地区：[1]东北林业大学信息与计算机工程学院,哈尔滨150040 [2]哈尔滨工业大学经济与管理学院,哈尔滨150001

出　　处：《黑龙江大学自然科学学报》2014年第2期250-254,共5页Journal of Natural Science of Heilongjiang University

基　　金：黑龙江省留学归国人员基金资助项目(LC2012C06)

摘　　要：有效地区分僵尸粉和真实用户对于评估企业营销活动的实际效果以及评价名人、网站等用户的真实影响力具有十分重要的价值。僵尸粉的形态已从过去以关注为主的低级形态向进行营销的高级形态过渡,为了有效地鉴别微博中的僵尸粉,本文通过构建用户的粉丝数、关注数、微博数、转发数、微博的转发情况、微博的评论情况、分时段的发博数等代表用户行为的特征向量,提出了一种基于用户行为的智能识别方法。通过实验证实,对比微博用户静态特征的识别方法,利用本文构建的基于用户行为的僵尸粉识别方法,准确率有较大提升。Spammer detection is very important for enterprise to measure the influence of marketing, influential individual and celebrity, which can distinguish the spammers from the normal ones. With the evolution of the spammer, it is not possible to identify spammers by using simple rules. Then, a classifier has been built using the machine learning algorithm combined with the features selected from the user profile and tweets, such as the number of followers, the number of retweets, and so on. The experimental results show that this classifier can achieve a higher accuracy than the ones usin~ the static features.

关 键 词：支持向量机 僵尸粉 数据挖掘 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]