网络社区动态角色挖掘算法研究*

传统的社区挖掘以社区为单位,忽略了社区内部成员的性质和地位。为了提高社区挖掘的精度,为个性化推荐提供一个优化的基础平台,基于优先情节和增长定律,提出了一种新颖的动态角色挖掘算法。首先根据节点的度数分布逆向推导社会网络的形成演化机制,构造网络时间轴;然后根据时间轴逐步向网络中添加新节点,同时进行社区挖掘和角色划分。在人工网络和真实世界网络上进行了多次测试,并与G-N算法进行了比较,取得了较好的结果。实验证明,应用动态角色挖掘算法得到的社区都是强连通社区,具有较高的准确性和实用价值。

Dynamic role assorted discovery of network community

Traditional community discovery algorithms study network members in community groups,while ignoring the inner members’ characteristics and status.In order to improve the discovery accuracy and provide an optimized platform for personalized recommendation system,this paper put forward a novel dynamic community discovery algorithm on the basis of complex priority and the grouth theorem.It firstly derived the social network’s mechanism of formation and evolution in negative direction according to the node-degree,at the same time,it constructed the time axis,then gradually put nodes into the network and divided them into different communities and gave distinct roles.It tested the algorithm on both artificial networks and real world networks,and compared with G-N.Experimental results show that,the discovery community are all strong connected communities and this algorithm has great pratical value as well as high accuracy.