PreSTyDe: Improving the Performance of within-project Defects Prediction by Learning to Classify Types of Software Faults

Gabriela Czibula, Ioana-Gabriela Chelaru, Arthur Molnar, Istvan Gergely Czibula

2024

Abstract

Software defect prediction (SDP) is an important task within software development. It is a challenging activity, as the detection of software modules that are prone to malfunction in new versions of software contributes to an improved testing process and also increases the quality of the software. In this paper, we propose a two-stage hybrid approach for predicting the error-proneness of the application classes in an upcoming version of a software project by employing a taxonomy of defects unsupervisedly uncovered from the previous software releases. The first stage of the proposed approach consists of an unsupervised labelling of software defects from the available versions of the analysed software system. During the second stage, a supervised classifier is used to predict the error proneness during the software project’s evolution employing the taxonomies of defects uncovered in the previous stage. Experiments carried out with Calcite software in a SDP scenario within a project highlighted that the performance of predicting software defects during a project evolution increases by approximately 5%, in terms of the average Area under the Receiver Operating Characteristic curve, by developing predictors for different classes of software defects.

Paper Citation

in Harvard Style

Czibula G., Chelaru I., Molnar A. and Gergely Czibula I. (2024). PreSTyDe: Improving the Performance of within-project Defects Prediction by Learning to Classify Types of Software Faults. In Proceedings of the 19th International Conference on Evaluation of Novel Approaches to Software Engineering - Volume 1: ENASE; ISBN 978-989-758-696-5, SciTePress, pages 214-225. DOI: 10.5220/0012732300003687

in Bibtex Style

@conference{enase24,
author={Gabriela Czibula and Ioana-Gabriela Chelaru and Arthur Molnar and Istvan Gergely Czibula},
title={PreSTyDe: Improving the Performance of within-project Defects Prediction by Learning to Classify Types of Software Faults},
booktitle={Proceedings of the 19th International Conference on Evaluation of Novel Approaches to Software Engineering - Volume 1: ENASE},
year={2024},
pages={214-225},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012732300003687},
isbn={978-989-758-696-5},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 19th International Conference on Evaluation of Novel Approaches to Software Engineering - Volume 1: ENASE
TI - PreSTyDe: Improving the Performance of within-project Defects Prediction by Learning to Classify Types of Software Faults
SN - 978-989-758-696-5
AU - Czibula G.
AU - Chelaru I.
AU - Molnar A.
AU - Gergely Czibula I.
PY - 2024
SP - 214
EP - 225
DO - 10.5220/0012732300003687
PB - SciTePress