微博网络上的重叠社群发现与全局表示

发布时间：2018-04-24 13:45

本文选题：微博网络 + 实体关系模型　；参考：《软件学报》2014年12期

【摘要】：微博网络是新兴的覆盖海量用户、涉及广泛话题并具有复杂重叠社群结构的多模网络.在深入研究微博网络各类实体和属性内在联系的基础上,提出了以用户-话题关系为主要划分原则的重叠社群表达模型及相应的社群结构发现算法.该方法不仅考虑网络中的用户-话题关系,还融合了这一网络特有的用户关注关系、博文评论与转发关系等所形成的复合网络关系.同时,改进了传统的社群隶属矩阵表述模型,通过引入虚拟社群,使隶属矩阵不仅合理反映个体对社群的隶属度,同时标识了个体在社群中的核心度.通过基于新浪微博数据集的实验验证,结果表明:该模型与方法能够高效合理地刻画该数据集包含的重叠社群结构,实验结果具有良好的可解释性,所提出的模型和算法可以有效地应用于类似多模网络社群划分和演化分析研究中.
[Abstract]:Weibo network is a new multi-mode network which covers a large number of users. It covers a wide range of topics and has a complex overlapping community structure. On the basis of deeply studying the internal relations between various entities and attributes of Weibo network, an overlapping community representation model based on user-topic relationship is proposed, and the corresponding community structure discovery algorithm is proposed. This method not only considers the user-topic relationship in the network, but also combines the complex network relationship formed by the unique user concern relationship, blog comment and forwarding relationship. At the same time, the traditional representation model of community membership matrix is improved. By introducing virtual community, the membership matrix not only reflects the membership degree of individual to community reasonably, but also identifies the core degree of individual in community. The experimental results based on Weibo data set of Sina show that the model and method can describe the overlapping community structure of the data set efficiently and reasonably, and the experimental results are good interpretable. The proposed model and algorithm can be effectively applied to the study of community partitioning and evolution analysis of similar multimode networks.
【作者单位】：南京大学计算机科学与技术系;淮海工学院计算机工程学院;
【基金】：国家自然科学基金(61403156,61375069,61105069) 国家博士后基金(2011M500846) 江苏省自然科学基金(11KJB520001,13KJB520002) 江苏省科技支撑计划(BE2012181)
【分类号】：TP393.092

【参考文献】