基于社交媒体的用户情绪建模与异常检测

摘要

For abnormal emotional detection among micro-blog users,this paper proposes ananomaly detection meth-od based on the joint probability density of multivariate Gaussian model and power-law distribution.In the experi-ments,the anomaly detection accuracy is 83.49% in terms of individual user,and 87.84% in terms of month.Sta-tistics reveals that individual users'neutral,happy and sad emotions fall into the normal distribution,but the amazed and angry emotions are not.Emotions of micro-blogs released by groups confirm to the power-law distribution,but not those by the individual.%为了对新浪微博用户的异常情绪进行检测和分析,该文提出一种基于多元高斯模型和幂律分布的异常检测方法,根据联合概率密度值判断用户是否出现情绪异常.在实验部分,按照不同用户的异常检测准确率为83.49%,按照不同月份为87.84%.分布测试表明,单个用户的中性、快乐和悲伤情绪服从正态分布,而惊讶和愤怒情绪则不服从;群体发布的微博的情绪服从"幂律分布",而单个用户则不服从.该文引入多元高斯模型来进行社交媒体的异常情绪的检测,通过联合概率密度值量化了异常情绪检测.当数据充足时,该方法可以检测用户或者某个社交平台每一周甚至每一天的异常情绪,这对个体异常情绪检测、网络舆情挖掘、大规模爆发事件预防以及公共安全监测有一定意义.