一种自动确定特征向量与类别数目的谱聚类算法

A spectral clustering algorithm for automatic determination of eigenvectors and category quantity

目前大多数应用于复杂网络社团划分的算法都不能自动确定类别数目.为了解决该问题,在计算出复杂网络的拉普拉斯矩阵特征值之后,利用特征值间隔提出一种自动确定特征向量与类别数目的谱聚类算法.该算法通过数据构造拉普拉斯矩阵,得到一系列特征值,再利用特征值差值确定聚类数目以及特征向量,然后利用K-means算法对特征向量进行处理最终得到社团划分结果.用MATLAB软件对常用网络进行测试,实验结果表明,该算法对复杂网络社团的划分具有较好的分类准确性.

Currently,most algorithms for community partition in complex network can not automatically determine the category quantity.In order to resolve this problem,a spectral clustering algorithm for automatic determination of eigenvector and category quantity is proposed by using the eigenvalue gap,which is obtained by computing the Laplacian eigenvalue of the complex network.In this algorithm,the Laplacian matrix is constructed by using the data to obtain a series of eigenvalues and determine the category and the Laplacian characteristic quantity and eigenvectors by using the difference of the eigenvalues.Then the K-means algorithm is used to process the eigenvectors and,finally,the result of community division is obtained.Some common networks are tested with software Matlab and the experimental result shows that this algorithm will have a better classification accuracy for the division of complex network community