一种基于复杂网络属性值的K-means聚类算法

传统-means聚类算法的性能依赖于初始聚类中心的选择.本文将复杂网络节点的属性值作为节点的度、聚集度与聚集系数的加权值,通过计算所有节点的加权综合聚集特征值,选取综合聚集特征值高,并且彼此之间无高聚集性特征的K个节点作为聚类的初始聚类中心,然后进行聚类迭代过程.实验结果表明,新算法对初始聚类中心的选取更迅速有效,避免了传统K-means算法初始聚类节点选取的敏感性,进而提高K-means算法的聚类质量.

A K-means cluster algorithm based on complex networks attribute value

Performance of the traditional K-means algorithm is dependent on the choice of the initial cluster centers.In this paper,the weighted values of the complex network nodes attributes are defined as the node degree,aggregation and clustering coefficient which are used to improve the initial cluster center selection of K-means algorithm.The aggregation characteristics of each node is calculated,and nodes with high aggregation values are merged and selected as initial cluster centers,then iteration is implemented for K-means clustering.Experimental results show that the new algorithm can select initial cluster centers effectively and isn’t sensitive to selecting process,quality of K-means clustering is improved.