Empirical Validation of Variable Method Interaction Cohesion Metric (VMICM) for Enhancing Reusability of Object-Oriented (O-O) Software

Abstract

Any object-oriented (O-O) module's primary goal is to build classes with a high level of coherent interaction between variables and methods. To increase the quality of O-O (Object-Oriented) software, various metrics emphasizing cohesiveness have been established so far. These metrics operate on both the design and the code levels. However, these metrics still fall short of fully measuring the cohesion of object-oriented (O-O) software. Based on several concepts of cohesive interlinkages between variables and procedures, the study proposed an enhanced cohesion metric. The four forms of cohesive linkages (VMRv, VMMv, VMRTv, and VMOv) between variables and procedures were the focus of this study. The axiomatic frame of reference was employed for theoretical validation, and univariate logistic regression was applied in the MATLAB environment for empirical validation. The approach of univariate logistic regression has been adopted because it provides incredibly accurate data and can even be applied to datasets that can be linearly separated. The proposed metric exhibits high cohesion, which is the ultimate perspective of a highly reusable Object- Oriented (O-O) module, as evidenced by the testing phase and even training the real dataset with reusability prediction in terms of high values of precision, recall, R2, and low value of RSME of VMICM metric. The study results demonstrated that the proposed metric can act as a measure for predicting the reusability of the Object-Oriented (O-O) system.