Empirical analysis for investigating the effect of object‐oriented metrics on fault proneness: a replicated case study

Abstract

Abstract The importance of software measurement is increasing, leading to the development of new measurement techniques. Many metrics have been proposed related to the various object‐oriented (OO) constructs like class, coupling, cohesion, inheritance, information hiding and polymorphism. The purpose of this article is to explore relationships between the existing design metrics and probability of fault detection in classes. The study described here is a replication of an analogous study conducted by Briand et al . The aim is to provide empirical evidence to draw the strong conclusions across studies. We used the data collected from Java applications for constructing a prediction model. Results of this study show that many metrics capture the same dimensions in the metric set, hence are based on comparable ideas and provides redundant information. It is shown that by using a subset of metrics prediction models can be built to identify faulty classes. The model predicts faulty classes with more than 90% accuracy. The predicted model shows that import coupling and size metrics are strongly related to fault proneness, confirming the results from previous studies. However, there are also differences reported in this study with respect to previous studies such as inheritance metric which counts methods inherited in a class is also included in the predicted model. Copyright © 2008 John Wiley & Sons, Ltd.