面向话题的微博网络测量研究  

Topic-oriented measurement of microblogging network

作　　者：刘玮[1,2,3] 王丽宏[3] 李锐光[3] 

机构地区：[1]中国科学院计算技术研究所 [2]中国科学院大学 [3]国家计算机网络应急技术处理协调中心

出　　处：《通信学报》2013年第11期171-178,共8页Journal on Communications

基　　金：国家自然科学基金资助项目(61170230);国家科技支撑计划基金资助项目(2012BAH46B01)~~

摘　　要：针对话题生成网络的动态时序特性,设计定量计算方法,从微博内容、网络结构、用户行为角度开展面向话题的新浪微博网络测量研究,结果发现:少数微博被大量转发,转发次数与对应微博数呈现近似的幂率分布;话题热度呈现明显的突发性和变化趋势,局部波动率能够有效地在大量背景微博中发现突发话题;基于话题生成的转发网络的小世界特性并不明显,且密集的关注关系不一定引发频繁的转发行为;传播能力强的话题中含有较大比例的持续参与用户,用户行为的话题相关性能够有效检测潜在关键用户。测量结果有助于了解话题生成网络的内容传播特点、网络结构特性及用户行为模式,测量指标能够有效应用于微博话题影响力分析等相关研究。According to the dynamic and temporal characteristics of the topic-generated network, a method of quantitative calculation was designed, and then the topic-oriented research on the network measurement technology from many aspects such as the features of the content was conducted, as well as the network topology and the characteristics of the user behavior. The experiments on the S1NA microblog showed four new results. The first is that only a small portion of tweets has been forwarded broadly and the number of retweets follows the power-law distribution. The second is that the tweets＇ number of one topic is episodic and changing frequently, and the burst topic can be detected by the local volatility feature found in the massive background microblog data. The third is that the small-world feature in the topic-generated retweeting network is not obvious, and the dense relationship doesn＇t necessarily induce the frequent retweeting behavior. The fourth is that the topic which has been propagated broadly usually has a large portion of the consistently participating users, and the correla- tion of the user behavior can be used to detect the potential and important users. The experimental results are helpful for un- derstanding the propagating mode, the structural characteristics and the pattern of the user behavior in a topic-generated net- work, and the indicators measured in the experiment can also be effectively applied in the future analyses.

关 键 词：内容特征 网络结构 用户行为 小世界特性 局部波动率 影响力分析 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]