Enhancing Software Fault Prediction Through Feature Selection With Spider Wasp Optimization Algorithm

Abstract

Software fault prediction (SFP) is a critical focus in software engineering, aiming to enhance productivity and minimize costs by detecting faults early. Feature selection (FS) is pivotal in SFP, enabling the identification of pertinent features for fault prognosis. Existing Feature Selection methods face challenges such as high computational complexity and poor generalization. This paper introduces Feature Selection using Spider Wasp Optimization (FSSWO), a novel FS approach employing the Spider Wasp Optimization (SWO) algorithm, specifically designed for SFP. FSSWO selects optimal feature subsets inspired by spider wasps’ behavior. The proposed FSSWO approach is compared with several existing feature selection algorithms, namely FS using Genetic Algorithm (FSGA), FS using Particle Swarm Optimization (FSPSO), FS using Differential Evolution (FSDE), and FS using Ant Colony Optimization (FSACO). Using eleven benchmark datasets, the performance of the proposed FSSWO technique has been assessed and contrasted with its equivalent. The results of the proposed FSSWO approach provide comparable and even superior results to the existing algorithms. The significance of the results has been statistically validated using Friedman and Holm tests. The statistical result of the proposed FSSWO approach reveals that the performance of proposed FSSWO models is improved which leads to better quality software at reduced costs.