•  
  •  
 

Document Type

Review

Abstract

The paper presents a detailed study of metaheuristic feature selection (MH-FS) methods in software fault prediction (SFP), designed to address the limitations of filter and wrapper feature selection methods such as high dimensionality and local optima that can lead to poor prediction accuracy. MH-FS algorithms address these issues by effectively searching the feature space to identify optimal subsets that improve classifier performance. It conducts a comprehensive narrative review and SWOT analysis (Strengths, Weaknesses, Opportunities and Threats) of MH-FS methods in SFP, considering 44 peer-reviewed articles from 2023 to 2025. A formal taxonomy of metaheuristic algorithms is presented, along with an extensive comparative analysis using classifiers, datasets and evaluation metrics. The review is based on clear research questions that address the challenges and difficulties of traditional feature selection methods, most commonly applied metaheuristic methods and also describes the frequently used classifiers that are available with MH-FS. Performance comparison is derived by evaluating the accuracy results reported in already existing studies that apply MH-FS algorithms with common classifiers and NASA datasets (KC1, JM1, CM1, and PC1). Furthermore, the paper identifies application domains, key research challenges and emerging directions in MH-FS models. This work provides a structured reference to researchers and practitioners for an informed algorithm selection process and to develop more robust, scalable and intelligent metaheuristic feature selection techniques for software fault prediction

Receive Date

13 Feb 2026

Revise Date

02 Sep 2026

Accept Date

06 Sep 2026

Publication Date

2026

Download

Share

COinS