Adopting six sigma in formulating a defect prediction model for system testing

Abstract

This paper describes the effort on applying Six Sigma approach in formulating a defect prediction model for system testing. The motivation behind this effort is to have an early quality indicator of the software entering system testing phase towards optimizing testing process, ensuring comprehensive test coverage, finding as many defects as possible and determining when to stop testing. By having such prediction, it is expected that all known defects could be contained within testing phase thus allows independent testing team to achieve zero known defects once software is delivered to end-user. Design for Six Sigma (DfSS) is used to build the model by relying on historical data to predict defects in system testing. Metrics from prior phases to system testing were collected and analyzed to identify the potential predictors and their relationships on defect occurrence in testing phase. Repeatability and capability of testers' consistency in discovering defects were also measured. Multiple regression analysis was performed using the collected metrics find out which combination of metrics has significant impact to the defect discovery. The mathematical equations were verified using new software projects to select the best model. The equation that produced the closest prediction against actual defects found was selected as the final defect prediction for system testing.