Toward software requirements modularization using hierarchical clustering techniques

Software modularity is not a new concept in the software engineering field; it has been a design issue since the earliest days of software development. Because the software designer cannot be expected to conceptualize a complex software application as a whole, it is usual to create a top-level design which is decomposed into a set of modules. The degree of modularization is a subjective concept that is difficult to measure; however, coupling and cohesion are two well-known concepts that are used to characterize software modularization.In this paper we illustrate how requirement scenarios can be clustered, based on attributes identified in the scenarios to quantitatively assess software modularization. Our technique uses a data mining clustering method that clusters scenarios, so that those scenarios within a cluster have a strong functional relationship with one another and weak relationships with scenarios in other clusters. Hence, cohesion within clusters is maximized while coupling between clusters is minimized. Consequently, software modularization based on these clusters should provide a good initial design.