DefinePK

Abstract

Identifying software defects has always been one of the major concerns of software designers and developers across the entire software industry. The intend of particular software that leads to produce some faults which is not determined previously during testing phase can lead to complete failure of the software and definitely incurred unnecessary cost and time inclusion. Thus the requirement of predicting faults at an earlier stage of its development has become necessity of today’s software development trend as the software becomes more distributed, complex and heavily cost infusive in nature. In this research work authors have proposed and implemented the idea of discretization of metric values in order to get better classification results. Authors have generated Boolean functions based on project metrics values, so that these values could be confined in the domain of classifying and predicting software faults. In this work authors have checked the performance of their proposed model by considering seventeen different software project and their version data taken from promise repository of NASA namely: Jedit, lucene, tomcat, velocity, xerces andxalan. The results gained after applying the proposed Boolean classifier are better and more promising in terms of its accuracy and precision, compared with available literature. The study claim visible increase in accuracy, as compared to other classifiers considered in the study like: Naïve Bayes, Random Forest, Perceptron, KNN and SOM.