ANALYSIS OF ENSEMBLE MODELS FOR AGING RELATED BUG PREDICTION IN SOFTWARE SYSTEMS

Abstract

With the evolution of the software industry, the growing software complexity led to the increase in the number of software faults. According to the study, the software faults are responsible for many unplanned system outages and a ects the reputation of the company. Many techniques are proposed in order to avoid the software failures but still software failures are common. Software fault prediction is the process, which predicts about the fault proneness of the software module. This process is based on some previous data (if available) and certain learning models. The prediction of the accurate location of faults can boost the testing process and allows the developers to focus on the critical modules that may account for the maximum number of faults. Many software faults and failures are outcomes of a phenomenon, called software aging. In this work, we have presented the use of various ensemble models for development of approach to predict the Aging Related Bugs (ARB). A comparative analysis of di erent ensemble techniques, bagging, boosting and stacking have been presented. The experimental study has been performed on the LINUX and MYSQL bug datasets collected from Software Aging and Rejuvenation Repository.