Applying heuristic approaches for predicting defect-prone software components

Abstract

Effective and efficient quality assurance has to focus on those parts of a software system that are most likely to fail. Defect prediction promises to indicate the defect-prone components of a software system. In this paper we investigate the viability of predicting defect-prone components in upcoming releases of a large industrial software system. Prediction models constructed with heuristic machine learning are used to classify the components of future versions of the software system as defective or defect-free. It could be shown that the accuracy of the predictions made for the next version is significantly higher (around 74%) than guessing even when taking only new or modified components into account. Furthermore, the results reveal that, depending on the specific prediction model, acceptable accuracy can be achieved for up to three versions in the future. © 2012 Springer-Verlag.