Applying heuristic approaches for predicting defect-prone software components

Authors Rudolf Ramler
Thomas Natschläger
Editors R. Moreno-Diaz
F. Pichler
A. Quesada
Title Applying heuristic approaches for predicting defect-prone software components
Booktitle Computer Aided Systems Theory - Proc. EUROCAST 2011, Part I
Type in proceedings
Publisher Springer
Series Lecture Notes in Computer Science
Volume 6927
ISBN 978-3-642-27548-7
Month February
Year 2012
Pages 384-391
SCCH ID# 1049
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.