Applications of Clustering to Early Software Life Cycle Phases

This paper presents empirical studies of applying numerical taxonomy to component clustering and software partitioning. The technique is based on cohesion and coupling information of a software system as these two properties have great impact on other software quality attributes such as maintainability, portability, scalability and reusability. The technique has been successfully applied to several real-time software systems in telecommunications and computer networks at various levels of abstraction or software life-cycle phases. Numerical taxonomy is mathematically simple and yet it provides a useful mechanism for component clustering and software partitioning. This paper presents experiments of the clustering techniques to various types of application and demonstrates a visualization tool that facilitates the analysis process.