基于依赖结构分析的类重要性度量方法  

Dependence structure analysis-based approach for measuring importance of classes

作　　者：周毓明[1,2] 徐宝文[1,3] 

机构地区：[1]东南大学计算机科学与工程学院,南京210096 [2]香港理工大学计算机系,中国香港 [3]江苏省软件质量研究所,南京210096

出　　处：《东南大学学报（自然科学版）》2008年第3期380-384,共5页Journal of Southeast University：Natural Science Edition

基　　金：国家自然科学基金资助项目(60425206,60633010)

摘　　要：为支持自动识别系统中重要的类,提出了一种基于软件依赖结构分析的类重要性度量方法.该方法首先用一个图模型表达类之间的3种依赖关系,包括类-类继承依赖、类-属性交互依赖和类-方法交互依赖,然后用PageRank,HITS和Betweenness Centrality算法通过对依赖关系结构的全局分析来度量类的重要性.该方法仅使用在软件开发的分析和设计阶段可以得到的信息,因而可用来在软件开发的早期识别系统中重要的类.实验结果表明,在识别重要的类时,该方法产生与简单方法不同的结果.To support the automatic identification of important classes in a system, a dependence structure analysis-based approach for measuring the importance of classes is proposed. This approach consists of two steps. First, a graph model is used to represent the dependence relationships between classes, including class-class inheritance dependence, class-attribute interaction dependence, and class-method interaction dependence. Then, PageRank, HITS （hyperlink-induced topic search）, and Betweenness Centrality algorithms are applied to measure the importance of classes by globally analyzing the dependence structure among classes. In this approach, only the information from the analysis and design stages of software development is needed. Therefore, it can be used to identify important classes at the early stages of software development. The experimental results show that, when used to identify important classes, the proposed approach produces different results from those obtained using simple approaches.

关 键 词：依赖结构 重要性 类 度量 面向对象 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]