Enhancing Object Oriented Coupling Metrics w.r.t. Connectivity Patterns

Abstract

Several Object Oriented metrics have been proposed in the literature. These metrics aim to measure the relationship between different classes and their methods. There are different ways in which classes are connected. Different metrics use different models to represent the relationship or connectivity pattern between the classes. In some cases, the Object Oriented metrics obtain the same value of coupling for different components having same models (i.e. same number of classes in a component) but different connectivity patterns. This leads incorrectly considering the components to be same in terms of coupling, even though their relationship or connectivity patterns clearly indicate that the degree of coupling are different. We refer this problem as Inability to Differentiate Anomaly (IDA). In this thesis we list and discuss Object- Oriented metric like Coupling between Objects (CBO), Response for Class (RFC) and Depth Inheritance Tree (DIT) in which the Inability to Differentiate Anomaly exists. We empirically study the frequent occurrence of IDA problem when the considered metrics are applied to different components with same class model. Finally, we propose a metric i.e. Coupling based on Strength Specification Metric (CSSM) which gives the distinct coupling value for components with same number of classes but different connectivity patterns. We compared and contrasted CBO and CSSMCBO, RFC and CSSMRFC, DIT and CSSMDIT and quoted the differences very clearly. And finally we give normalized CSSM which lie between the ranges 0 to 1 which classifies the components into three categories i.e. complex, medium and low. Metrics based on the strength of coupling parameters is more desirable because it exhibits a lower chance of incorrectly considering components to be equally coupled when they have different connectivity patterns.