结合熵有效性函数的FCM算法识别社团结构

FCM Combined with Validity Measure Function of Entropy to Identity Community Structure

挖掘和发现复杂网络中的社团结构是复杂网络研究的基础性问题。针对复杂网络中的社团结构往往具有重叠性,提出了结合熵有效性函数的模糊聚类(Fuzzy c-means,FCM)算法。首先基于信息熵提出了熵有效性函数,用于确定网络的"最佳"聚类数;其次给出了聚类数范围和两个过滤条件;最后将三者与FCM算法相结合,应用到Zachary’s karate club network、Dolphin social network和American college football network的社团结构检测。为了进一步体现熵有效性函数的优越性,将熵有效性函数和模块度函数,分别与k-means算法相结合,对3个网络进行了实验。实验结果表明,熵有效性函数可以较准确的找到"最佳"聚类数,且结合熵有效性函数的FCM算法划分结果精确度都在90%以上。

Mining and identifying community structure in complex networks is a fundamental issue in complex network research. In view of there are often overlapping commumtles In complex networK, ail improved Fuzzy c- means（FCM） method based on validity measure function of entropy is proposed. Firstly, devise a validity measure function of entropy, and use to determine the ＂best＂ clustering num- ber of a network. Secondly, point out a range of clustering number and two filter conditions. Finally, combine them with FCM to detect communities in Zachary＇s karate club network, Dolphin social network and American college football network. In order to further demonstrate the superiority of validity measure function of entropy effectiveness, combine validity measure function of entropy and modularity function with k- means method, respectively, on the three networks. Experimental results show that the validity measure function of entropy could accurately find the actual clustering number, and the improved FCM method＇s accuracy of clustering is above 90X.