基于凝聚式信息瓶颈的加权层次聚类算法

Weighted Hierarchical Clustering Algorithm Based on Agglomerative Information Bottleneck

提出一种针对面向对象软件架构恢复的基于凝聚式信息瓶颈的加权层次聚类算法(ABWHC)。该算法采用信息丢失度作为相似度度量标准,扩充聚类特征和权值,利用面向对象软件的特性,为实体或簇生成用以描述其含义的标签组。实验结果表明,ABWHC算法不仅能改善聚类的性能,还能恢复面向对象软件的架构。

This paper proposes an Agglomerative Information Bottleneck based Weighted Hierarchical Clustering algorithm（ABWHC） to rebuild the architecture of object oriented software.ABWHC uses information loss as the similarity measure,considers the characteristics of object oriented software by extending clustering features and weights,and generating label group for each entity or cluster.Experimental results demonstrate that ABWHC improves the performance of clustering,and efficiently and flexibly achieves object oriented software architecture recovery.