Classification Techniques Use to Empirically Validate Redundancy Metrics as Reliability Indicators based on Fault-proneness Attribute

Dalila Amara, Latifa Rabai

2022

Abstract

Software metrics are proposed as quantitative measures of internal quality factors like cohesion and complexity. For the external ones such as reliability and maintainability, they are usually predicted by means of various metrics of internal attributes. In this context, we have focused on a suite of four entropy-based software redundancy metrics considered as software reliability indicators. Despite their important purpose, they are manually computed and only theoretically validated. Hence, we have implemented an empirical approach for assessing these metrics, using a set of programs retrieved from real software projects. Given that software reliability as external attribute, cannot be directly evaluated, we employ other measurable quality factors representing direct reflections of this attribute. Among them, defect density and fault-proneness are widely used as means to measure and predict software reliability based on software metrics. The basic idea is to generate an empirical dataset embodying for each program, the values of the redundancy metrics and the values of one of these measurable attributes. In our previous work, we have studied their relationship with the defect density attribute in order to validate them as useful reliability indicators. Promising results indicating the usefulness of these metrics as defect density indicators are obtained. Classifying modules (functions or classes) as defective or not defective is also an important reliability indicator. Literature review shows that software reliability counts on its fault-prone modules and more trusted software consists of less fault-prone units. Therefore, we aim in this paper to propose an empirical approach to validate the redundancy metrics as significant reliability indicators. The validation is carried out using the accuracy measure and results show that the fault proneness attribute can be predicted using the redundancy metrics with a good accuracy rate of 0.82.

Paper Citation

in Harvard Style

Amara D. and Rabai L. (2022). Classification Techniques Use to Empirically Validate Redundancy Metrics as Reliability Indicators based on Fault-proneness Attribute. In Proceedings of the 17th International Conference on Evaluation of Novel Approaches to Software Engineering - Volume 1: ENASE, ISBN 978-989-758-568-5, pages 209-220. DOI: 10.5220/0011081900003176

in Bibtex Style

@conference{enase22,
author={Dalila Amara and Latifa Rabai},
title={Classification Techniques Use to Empirically Validate Redundancy Metrics as Reliability Indicators based on Fault-proneness Attribute},
booktitle={Proceedings of the 17th International Conference on Evaluation of Novel Approaches to Software Engineering - Volume 1: ENASE,},
year={2022},
pages={209-220},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0011081900003176},
isbn={978-989-758-568-5},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 17th International Conference on Evaluation of Novel Approaches to Software Engineering - Volume 1: ENASE,
TI - Classification Techniques Use to Empirically Validate Redundancy Metrics as Reliability Indicators based on Fault-proneness Attribute
SN - 978-989-758-568-5
AU - Amara D.
AU - Rabai L.
PY - 2022
SP - 209
EP - 220
DO - 10.5220/0011081900003176