基于节点相似性的LFM社团发现算法

摘要

In network with fuzzy community structure,precision of the traditional LFM algorithm decreases apparently.In order to solve this problem,an LFMJ algorithm is presented.Using the information of neighbor nodes and improved Jaccard coefficient,this algorithm reconstructed the network structure,and improved the precision of community division results.To validate the algorithm,five algorithms was tested in LFR benchmark and real networks,including LFMJ,traditional LFM,LPA algorithm and WT,FUA algorithm,which have better performance in com munity detection.The results show that,in LFR network,the accuracy of LFMJ is higher than both LFM and LPA,equaling to WT and FUA algorithm.In real network and LFR network with overlapping community,LFMJ gets the highest accuracy than others.The effectiveness of the algorithm is proved.%传统的局部适应度社团发现算法(LFM)在社团结构模糊的网络中精度下降严重.针对此问题,提出LFMJ算法.利用邻居节点信息和改进的杰卡德系数重构网络,使网络结构更为清楚,社团划分结果更为准确.为验证算法,选择了5种算法在LFR网络和真实网络中进行测试,包括LFMJ、LFM和传统的LPA算法以及性能较好的WT和FUA算法.结果表明:在标准LFR网络中,LFMJ精度高于LFM和LPA,与FUA和WT相当;在真实网络和具有重叠结构的LFR网络中,LFMJ精度优于其他4种算法.