基于模块性指标优化的层次聚类算法

层次形成的正确性决定了层次聚类的质量，通常围绕对象类内类间关系评价实现。本文基于聚类目标，综合考虑类内类问关系，借鉴网络分析中模块性评价准则，设计用于层次聚类的模块性指标，并采用自底向上合并的途径实现指标优化从而完成聚类，提出一种基于模块性指标优化的层次聚类算法。仿真试验表明，和谱聚类算法相比，本文介绍的算法实现简单，能以较少的计算代价，准确地获得样本特征，实现聚类。

Hierarchical Clustering Algorithm Based on the Modularity Criterion Optimization

The accuracy of forming hierarchies is very important to hierarchical clustering,which is achieved usually around the evaluation of relations among and inside object clusters. Based on the clustering essence and comprehensive consideration of intra-cluster and inter-cluster relation, modularity evaluation criterion in network analysis is adopted for reference to design the clustering modularity criterion function, and the criterion is optimized through sub-classes combination from bottom to top so as to implement clustering. According to this approach, a hierarchical clustering algorithm based on the modularity criterion optimization is developed. Simulation results suggest that comparing to spectral clustering algorithms,our algorithm is more straightforward,capable to extract more accurate critical features of the samples and then accomplish clustering with less computational cost.