基于链路预测的社会网络事件检测方法

发布时间：2018-01-29 12:27

本文关键词： 社会网络分析事件检测链路预测网络演化分析网络波动性分析　出处：《软件学报》2015年09期 　论文类型：期刊论文

【摘要】：网络演化分析与事件检测,是当前社会网络研究的热点和难点.现有的研究工作主要是针对网络提出不同的模型,并用网络特征指标对仿真结果进行评价.这些方法存在如下问题:(1)每种方法仅针对特定网络,通用性不高;(2)特征指标多种多样,不同模型的表现情况缺乏统一的评价标准;(3)未考虑网络演化的时间特性,难以描述网络演化的波动性,无法检测事件.针对上述问题,提出一种基于链路预测的社会网络事件检测方法 Link Event(由相似性计算算法Sim C和事件检测算法Event D组成),它可以对不同网络的波动性进行统一评价,并依此建立事件检测模型.主要工作包括:(1)证明了链路预测可以反映网络演化机制,相同机制下的模型演化法和链路预测在分析网络演化上具有内在的一致性;(2)基于链路预测,提出一种网络相似性计算算法Sim C(similar computing),并在考虑微观因素的基础上进行改进;(3)利用相似性计算结果,提出一种事件检测算法Event D(event detecting)检测出新事件.在不同特征的网络上进行实验,结果表明:所提出的Link Event方法能够较好地解决网络演化波动性问题,实现事件检测;同时也证明了利用链路预测技术进行网络演化分析的可行性以及相似性计算和事件检测算法的有效性.
[Abstract]:Network evolution analysis and event detection are the hot and difficult points in the current social network research. The existing research work is mainly to propose different models for the network. The simulation results are evaluated with the network characteristic index. These methods have the following problem: 1) each method is only for a specific network, so it is not universal. (2) the characteristic indexes are various, and the performance of different models is lack of uniform evaluation criteria; It is difficult to describe the volatility of network evolution and detect events without considering the time characteristics of network evolution. A social network event detection method based on link prediction (Link event) is proposed, which consists of similarity calculation algorithm (Sim C) and event detection algorithm (Event D). It can uniformly evaluate the volatility of different networks and establish an event detection model. The main work includes: 1) it proves that link prediction can reflect the evolution mechanism of the network. The model evolution method and link prediction under the same mechanism have inherent consistency in analyzing network evolution. (2) based on link prediction, a network similarity calculation algorithm, Sim similar computing algorithm, is proposed and improved on the basis of considering microcosmic factors. 3) based on the similarity calculation results, a new event detection algorithm, Event event detection algorithm, is proposed to detect new events. Experiments are carried out on networks with different features. The results show that the proposed Link Event method can solve the problem of network evolution volatility and realize event detection. At the same time, the feasibility of using link prediction technology to analyze network evolution and the effectiveness of similarity calculation and event detection algorithms are also proved.
【作者单位】：武汉大学计算机学院;软件工程国家重点实验室(武汉大学);
【基金】：国家自然科学基金(70901060,61471274) 湖北省自然科学基金(2011CDB461) 软件工程国家重点实验室(武汉大学)开放基金(SKLSE 2010-08-15) 武汉市科技局青年晨光计划(201150431101) 武汉市科技重大计划项目(2015010101010023)
【分类号】：TP393.02
【正文快照】： 社会网络是不断发展变化的,网络演化分析与事件检测是社会网络分析的重要组成部分.网络演化分析是指通过跟踪网络不同阶段的特征变化来描述其演化规律,进而分析网络增长、传播等行为,预测未来结构,甚至加以人为干预,以得到预期结果.网络演化分析技术已经随着社交网络的爆炸式

【参考文献】