基于流量时间序列的社交网络事件聚类分析  

Clustering analysis of social network events based on traffic time series

作　　者：刘静 张鸽 LIU Jing;ZHANG Ge(School of Software,Shanxi Agricultural University,Taigu Shanxi 030801,China)

机构地区：[1]山西农业大学软件学院,山西太谷030801

出　　处：《智能计算机与应用》2023年第9期164-167,共4页Intelligent Computer and Applications

基　　金：山西省重点实验室开放基金(CICIP2021005)。

摘　　要：社交网络上用户发布的在线内容非常不稳定,用户对每个事件的关注度也随时变化。虽然每个事件的关注程度各不相同,但某些具有共同特征的事件会呈现出相似的流量模式,本文旨在根据社交网络事件的流量时间序列对事件进行聚类,找到事件的共性特征。首先,利用皮尔逊相关系数来确定各事件的主题标签;然后,利用各事件的主题标签获得每隔固定时间有关该事件的推文总量,即该事件的流量时间序列;最后,利用K-SC(K-Spectral Centroid)聚类算法对事件的流量时间序列进行聚类,并分析聚类结果中每一类事件的共性特征。利用推特上2020东京奥运会期间场地自行车比赛事件的推文数据,验证了本文方法对基于流量时间序列的社交网络事件进行聚类分析的有效性。The online content posted by users on social networks is highly unstable,and users'interest in each event varies over time.Although the level of attention differs for each event,some events with common characteristics exhibit similar traffic patterns.This paper aims to cluster the events of social network based on their traffic time series to identify common features among them.Specifically,Pearson correlation coefficient is used to determine the thematic labels of each event.Then the total number of tweets related to each event at fixed intervals are calculated to represent the traffic time series of each event.Finally,the K-SC clustering algorithm is employed to cluster the traffic time series of the events,and the common features of events in the same cluster are analyzed.In the experiments,Twitter data of track cycling races during the 2020 Tokyo Olympics are used to validate the effectiveness of the proposed method in clustering social network events based on traffic time series.

关 键 词：社交网络 流量模式 流量时间序列 K-SC聚类算法 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]