Predictive Power of Two Data Flow Metrics in Software Defect Prediction

Adam Roman, Rafał Brożek, Jarosław Hryszko

2023

Abstract

Data flow coverage criteria are widely used in software testing, but there is almost no research on low-level data flow metrics as software defect predictors. Aims: We examine two such metrics in this context: dep- degree (DD) proposed by Beyer and Fararooy and a new data flow metric called dep-degree density (DDD). Method: We investigate the importance of DD and DDD in SDP models. We perform a correlation analysis to check if DD and DDD measure different aspects of the code than the well-known size, complexity, and documentation metrics. Finally, we perform experiments with five different classifiers on nine projects from the Unified Bug Dataset to compare the performance of the SDP models trained with and without data flow metrics. Results: 1) DD is noticeably correlated with many other code metrics, but DDD is not correlated or is very weakly correlated with other metrics considered in this study; 2) both DD and DDD are highly ranked in the feature importance analysis; 3) SDP models that use DD and DDD perform better than models that do not use data flow metrics. Conclusions: Data-flow metrics: DD and DDD can be valuable predictors in SDP models.

Paper Citation

in Harvard Style

Roman A., Brożek R. and Hryszko J. (2023). Predictive Power of Two Data Flow Metrics in Software Defect Prediction. In Proceedings of the 18th International Conference on Evaluation of Novel Approaches to Software Engineering - Volume 1: ENASE, ISBN 978-989-758-647-7, SciTePress, pages 114-125. DOI: 10.5220/0011842200003464

in Bibtex Style

@conference{enase23,
author={Adam Roman and Rafał Brożek and Jarosław Hryszko},
title={Predictive Power of Two Data Flow Metrics in Software Defect Prediction},
booktitle={Proceedings of the 18th International Conference on Evaluation of Novel Approaches to Software Engineering - Volume 1: ENASE,},
year={2023},
pages={114-125},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0011842200003464},
isbn={978-989-758-647-7},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 18th International Conference on Evaluation of Novel Approaches to Software Engineering - Volume 1: ENASE,
TI - Predictive Power of Two Data Flow Metrics in Software Defect Prediction
SN - 978-989-758-647-7
AU - Roman A.
AU - Brożek R.
AU - Hryszko J.
PY - 2023
SP - 114
EP - 125
DO - 10.5220/0011842200003464
PB - SciTePress