Title: Feature selection using Tasmanian devil optimisation algorithm for software fault prediction

Authors: Himansu Das; Muskan; Hrishikesh Kumar

Addresses: School of Computer Engineering, KIIT University, Bhubaneswar, India ' School of Computer Engineering, KIIT University, Bhubaneswar, India ' School of Computer Engineering, KIIT University, Bhubaneswar, India

Abstract: Software quality is improved through early software flaw identification, which is made possible by software fault prediction (SFP). Early software problem identification reduces the amount of money and work needed to fix those vulnerabilities. This paper presents a novel FSTDO model, a feature selection (FS) approach built upon Tasmanian devil optimisation (TDO). The proposed FSTDO helps to select the optimal feature subset that yields maximum classification accuracy. The proposed FSTDO algorithm is juxtaposed with four well-known FS algorithms, such as FS using ACO, FS using GA, FS using PSO and FS using DE. Four popular classifiers KNN, NB, DT, and QDA, were used with the FSTDO to assess the quality of its outcomes. The experimental outcomes reveal that the proposed FSTDO has more potential in comparison with the competing FS models as it delivers enhanced performance in terms of accuracy, and the selection of optimal features is also significantly viable.

Keywords: software fault prediction; SFP; feature selection; optimisation; Tasmanian devil optimisation; TDO; machine learning; classifiers; ant colony optimisation; ACO; particle swarm optimisation; PSO; K-nearest neighbour; KNN.

DOI: 10.1504/IJCSE.2025.143468

International Journal of Computational Science and Engineering, 2025 Vol.28 No.1, pp.32 - 55

Received: 12 Aug 2023
Accepted: 23 Jan 2024
Published online: 21 Dec 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article