Software similarity-based functional cohesion metric

Abstract

Cohesion is an important factor used in evaluating software design quality and modularity. The cohesion of a module refers to the relatedness of the module components. In software engineering, highly cohesive modules are highly desirable because of their high reusability and maintainability. Cohesion is classified according to levels. Functional cohesion, the strongest level, refers to how closely the module parts that contribute to different outputs are related. Here, a similarity-based functional cohesion (SBFC) metric is introduced to measure the functional cohesion of a module in a procedural or object-oriented program. The metric uses the degree of similarity between the data slices of the module as a basis to measure functional cohesion. The appropriateness of the metric is evaluated both theoretically and empirically. The evaluation results show that the metric does as well as some earlier metrics in indicating the level of cohesiveness and it does better than some in terms of providing different values for the modules of different cohesion. In addition, the SBFC metric is used as an indicator for restructuring the weakly cohesive modules.