An Automatic kernel of Graph Clustering Method in Conforming Clustering Number

摘要

Based on analyzing graph theory knowledge and kernel function theory,every data sample is considered as top point V in graph,so all data samples consist of nondirectional weighted graph G = ,which takes similarity as weighted value.In the perspective of graph theory,this article defines connected modulus,which can fully reflect the best clustering number.This modulus categorizes similar text into a connected graph,and keeps the clearance of physical meaning.In this paper,a Kernel of Graph Clustering method based on clustering was proposed,this arithmetic is compared with kernel C-equal value arithmetic.The test justifies that this arithmetic not only has less complexness in time and space,but also good robustness.