基于异同性的社区演化分类方法

The Community Evolution Classification Method on Similarity and Difference

动态网络中的社区演化分析是目前的研究热点之一,其在舆论控制、网络营销和个性化推荐服务等方面有着重要作用。提出一种基于节点重要性评价指标的差值吸收核心节点检测算法,首先计算各节点的相对权重值,进而划分核心节点,并以此为基础优化差异性公式,提出一种异同性社区演化分类模型,从相似性和差异性两方面对演化类型进行划分。将提出的分类模型与GED及SGCI在HEP-TH和波兰政治博客圈数据集上进行比较,实验结果表明,提出的分类模型在整体上优于GED及SGCI,尤其在Forming和Dissolving事件的检测时,可以做到对小社区敏感,能检测到小社区的多种演化类型。

The analysis of community evolution in dynamic network is one of the current research hotspots,which plays an important role in public opinion control,network marketing,personalized recommendation service and so on.This paper proposes a nonparametric core node detection algorithm based on the evaluation index of node importance,which is used to find the core nodes in the community nodes to form the core node set,and to judge the difference of community based on the core node set.Based on this,a classification model of community evolution based on similarities and differences is proposed.This model determines the evolution relationship and divides the evolution type from the two aspects of similarities and differences.Comparing the proposed classification model with GED and SGCI in hep-th and Polish political blogosphere data sets,the experimental results show that the proposed classification model is better than GED and SGCI classification model on the whole.Especially in the detection of forming and dissolving events,it can be sensitive to small communities,and can detect a variety of evolution types of small communities.