A clustering algorithm for identifying reusable system components

In the creation of computer based systems, demands for increased complexity and reduced timescales and costs are putting great strain on current development methods. One way of tackling this problem is increased reuse At a technical level the most promising approach is Product Line Development in which the software development process is separated into Application Family Engineering which aims to create reusable components, and Systems Engineering, which builds systems using those components. The problem is mapping the required functionality onto components. This paper presents a new technique for the construction of the model and components. The method starts with user level requirements from systems in an Application Family. Application Family level iscriminants are identified and a model is built in which these discriminates are represented in a signed graph. A clustering algorithm is then used to identify reusable components. A case study is presented where the method has been applied