Enhancing the fault prediction accuracy of CK metrics using high precision cohesion metric

Abstract

Object-oriented programs can be viewed as a collection of objects communicating with each other to perform a unique task. Many complex commercial applications have taken the object-oriented approach because of the benefits that it offers. The need for a reliable software resulted in the study and analysis of object-oriented metrics. The Chidamber and Kemerer (CK) metric suite has been considered as a pioneering work on object-oriented metrics and is the default standard for any new metric to be compared against. In this paper we evaluate the fault prediction capability of CK metric suite and validate it empirically. To further improve the accuracy of fault prediction we explore replacing the cohesion metric (LCOM) in CK suite with the proposed cohesion metric (high precision cohesion metric). We have considered data from 500 classes spread across 12 projects for the study. The results show that there is a considerable improvement in the prediction accuracy.