关键词: 重构风险, 软件模块化, 程序聚类, 软件度量, 软件质量

Abstract: As software evolves, its modularity gradually degrades. Software refactoring is an important means for software modularity adjustment, but which modules are most in need of refactoring is difficult to predict. This paper proposes a novel approach for measuring the refactoring risk of modules by software clustering. Using structural clustering and semantic clustering, two different kinds of implied modularity views can be recovered from the implementation as reference modularity. By comparing the differences between the realistic modular structure and the reference modular structure, modules with high refactoring risk are identified. A comparative experiment conducted on three open source softwares shows that the predicting result produced by the proposed approach conforms well to the actual refactoring activities, indicating the effectiveness of the proposed approach.

Key words: refactoring risk, software modularity, software clustering, software metrics, software quality