Decomposing object-oriented class modules using an agglomerative clustering technique
Authors:
- 1 September 2009
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 10636773,p. 93-101
Abstract
Software can be considered a live entity, as it undergoes many alterations throughout its lifecycle. Furthermore, developers do not usually retain a good design in favor of adding new features, comply with requirements or meet deadlines. For these reasons, code can become rather complex and difficult to understand. More particularly in object-oriented systems, classes may become very large and less cohesive. In order to identify such problematic cases, existing approaches have proposed the use of cohesion metrics. However, while metrics can identify classes with low cohesion, they cannot identify new or independent concepts. Moreover, these methods require a lot of human interpretation to identify the respective design flaws. In this paper, we propose a class decomposition method using an agglomerative clustering algorithm based on the Jaccard distance between class members. Our methodology is able to identify new concepts and rank the solutions according to their impact on the design quality of the system. Finally, our method has been evaluated by two independent designers who were asked to comment on the suggestions produced by our technique on their projects. The designers provided feedback on the ability of the method to identify new concepts and improve the design quality of the system in terms of cohesion.Keywords
This publication has 17 references indexed in Scilit:
- Diagnosing Design Problems in Object Oriented Systems2006
- Spectral and meta-heuristic algorithms for software clusteringJournal of Systems and Software, 2005
- The weighted combined algorithm: a linkage algorithm for software clustering2004
- Refactoring - improving coupling and cohesion of existing code2004
- Metrics based refactoring2003
- Experiments with clustering as a software remodularization method2003
- Automatic clustering of software systems using a genetic algorithm2003
- Software botryology. Automatic clustering of software systems2002
- Using automatic clustering to produce high-level system organizations of source code2002
- Component clustering based on maximal association2002