一种基于层次聚类的软件架构恢复方法

Hierarchical Clustering Based Software Architecture Recovery Approach

针对软件聚类侧重相似度测度而欠缺考虑实体和特征的特性的问题,提出一种基于层次聚类的软件架构恢复方法(HCSAR)。该方法有针对性地选取实体和特征,提出特征的多重加权策略,采用信息丢失度作为相似度测度,选取和设计软件聚类的客观和主观评估准则。与目前效果较好的软件聚类方法相比,HCSAR在聚类中期能生成更多的簇,主观判定数更低,能够通过调整关注点获得不同的聚类结果,使用设计的评估准则分析聚类结果还能有效辅助系统划分。

To solve the problem of the lack of consideration of the characteristics of entities and features,a hierarchical clustering based software architecture recovery approach was proposed.Targeted entities and features were selected,weight scheme was proposed to generate feature vectors,information loss was considered as the similarity metric,and objective and subjective evaluation measures were respectively chosen and given.Compared with superior agglomerative software clustering approaches,HCSAR is more cohesive,requires less arbitrary decisions,is flexible to adjust the focus of software clustering,and is able to assist to generate more accurate system partition.