Logo image
Back
A new binary chaos-based metaheuristic algorithm for software defect prediction
Journal article   Open access   Peer reviewed

A new binary chaos-based metaheuristic algorithm for software defect prediction

Bahman Arasteh, Keyvan Arasteh, Ali Ghaffari and Reza Ghanbarzadeh
Cluster computing, Vol.27(7), pp.10093-10123
01/10/2024
DOI: https://doi.org/10.1007/s10586-024-04486-4
Appears in  Recent Faculty of Science and Engineering Publications
pdf
A new binary chaos-based metaheuristic algorithm for software defect prediction1.60 MBDownloadView
Published (Version of record)CC BY V4.0,  Open Access
url
A new binary chaos-based metaheuristic algorithm for software defect predictionView
Published (Version of record)CC BY V4.0,  Open

Related links

Metrics

2 File views/ downloads
30 Record Views
19 Times Cited - Web of Science

Abstract

Software defect prediction Binary olympiad optimisation algorithm Feature selection Machine learning Module classification
Software defect prediction is a critical challenge within software engineering aimed at enhancing software quality by proactively identifying potential defects. This approach involves selecting defect-prone modules ahead of the testing phase, thereby reducing testing time and costs. Machine learning methods provide developers with valuable models for categorising faulty software modules. However, the challenge arises from the numerous elements present in the training dataset, which frequently reduce the accuracy and precision of classification. Addressing this, selecting effective features for classification from the dataset becomes an NP-hard problem, often tackled using metaheuristic algorithms. This study introduces a novel approach, the Binary Chaos-based Olympiad Optimisation Algorithm, specifically designed to select the most impactful features from the training dataset. By selecting these influential features for classification, the precision and accuracy of software module classifiers can be notably improved. The study's primary contributions involve devising a binary variant of the chaos-based Olympiad optimisation algorithm to meticulously select effective features and construct an efficient classification model for identifying faulty software modules. Five real-world and standard datasets were utilised across both the training and testing phases of the classifier to evaluate the proposed method's effectiveness. The findings highlight that among the 21 features within the training datasets, specific metrics such as basic complexity, the sum of operators and operands, lines of code, quantity of lines containing code and comments, and the sum of operands have the most significant influence on software defect prediction. This research underscores the combined effectiveness of the proposed method and machine learning algorithms, significantly boosting accuracy (91.13%), precision (92.74%), recall (97.61%), and F1 score (94.26%) in software defect prediction.

Details

Logo image