一种新的层次谱聚类算法

New Hierarchical Spectral Clustering Algorithm

提出一种新的聚类算法——层次谱聚类算法.该算法在传统二分的SM谱聚类的过程中嵌入了层次聚类算法,目的是为了提高谱聚类的聚类正确率,同时又利用谱聚类纠正了层次聚类过程中所得到的歪斜划分.实验结果表明:提出的层次谱聚类算法的聚类正确率比层次聚类算法、谱聚类算法的聚类正确率都要高,同时又纠正了层次聚类过程中的歪斜划分.

A new clustering algorithm--hierarchical spectral clustering algorithm was proposed. In the algorithm the hierarchical clustering was embedded in traditional dichotomy SM spectral clustering, in order to raise the accuracy of spectral clustering and check skewed divisions in the process of hierarchical clustering. The experimental results show that the hierarchical spectral clustering is superior to spectral clustering or hierarchical clustering on clustering accuracy and checking skewed divisions in the process of hierarchical clustering.