Three-way decision based Defect Prediction for Object Oriented Software

Abstract

Early prediction of defective software module plays critical role in the software project development to reduce the overall development time, budgets and increases the customer satisfaction. The bug prediction based on two-way classification method classifies the software module as defective or non-defective. This method provides good accuracy measure but this metric is not sufficient in case if misclassification cost is concerned. Classifying the defective module as non-defective will lead to higher cost of entire software project at the end. In this study, three-way decision based classification method and Random Forest ensemble are used to predict the defect in Object Oriented Software to reduce the misclassification cost which will lead to avoid the cost overrun. The eclipse bug prediction dataset is used and experimental results show that the decision cost is reduced and accuracy is increased using our proposed method.