Software defect prediction using process metrics elasticsearch engine case study

New technology advances provide more sophisticated features which increasingly require complex software. The growing complexity of software presents a significant challenge to software quality and reliability. The prediction of defects in software is aimed at maintaining software quality without allocating a lot of resources in the quality assurance activities. The distribution of defects in modules is not uniform. Defect prediction models can locate the most defective modules. In this study, software defect prediction was conducted using the source code of the ElasticSearch engine. The Logistic Regression, Naïve Bayes and J48 techniques were used to learn from previous versions and predict defects in upcoming releases. The J48 outperformed the Logistic Regression and Naïve Bayes. The quality and accuracy of the prediction models was evaluated using basic performance measures and the 10-fold cross validation.