Regression Tree Modeling for the Prediction of Software Quality

This paper demonstrates the use of regression tree models to predict the number of faults in a software module based on the software complexity metrics, prior to the testing phase, which can help in channeling the validation and testing efforts in a productive direction. We also compare the regression tree modeling technique with the fault density technique which is a very commonly used approach to predict the number of faults. 1. Introduction The production of modern computer software is one of the most complex and unpredictable activities in the industry. Software is a crucial part of many critical applications and certifying software integrity is absolutely essential. It is no longer acceptable to postpone the assurance of software quality until prior to a product's release. Delaying corrections until testing and operational phases can lead to higher costs, and it may be too late to improve the software product significantly. A software failure is a manifestation of a fault, which..