Statistical assessment of nonlinear manifold detection-based software defect prediction techniques

摘要

Prediction of software defects has immense importance for obtaining desired outcome at minimised cost and so attracted researchers working on this topic applying various techniques, which were not found fully effective. Software datasets comprise of redundant features that hinder effective application of techniques resulting inappropriate defect prediction. Hence, it requires newer application of nonlinear manifold detection techniques (nonlinear MDTs) that has been examined for accurate prediction of defects at lesser time and cost using different classification techniques. In this work, we analysed and tested the effect of nonlinear MDTs to find out accurate and best classification technique for all datasets. Comparison has been made between the results of without or with nonlinear MDTs and paired two-tailed T-test has been performed for statistical testing and verifying the performance of classifiers using nonlinear MDTs on all datasets. Outcome revealed that among all nonlinear MDTs, FastMVU makes most accurate prediction of software defects.