A Novel N-Shaped Transfer Function with a Behavior-Enhanced Mountain Gazelle Optimizer for Feature Selection

Abstract

This repository contains the source code, datasets, experimental results, and evaluation materials associated with the study of the Improved Mountain Gazelle Optimizer (IMGO) and the proposed N-shaped transfer function for continuous and binary optimization. The proposed IMGO enhances the original Mountain Gazelle Optimizer (MGO) by incorporating chaos-enhanced coefficient vectors and three life-inspired search strategies: male competition, ownership, and mating. These mechanisms are designed to improve population diversity, reduce premature convergence, and achieve a better balance between exploration and exploitation. The repository also includes the implementation of a novel N-shaped transfer function for converting continuous solutions into binary representations. The proposed transfer mechanism provides a non-monotonic mapping that supports adaptive continuous-to-binary conversion and is used in feature selection experiments. The provided materials cover the following experimental components: Continuous optimization experiments using the CEC 2017 benchmark functions to evaluate the performance of IMGO. Feature selection experiments using UCI datasets, where IMGO is applied with the N-shaped transfer function. Binary feature selection experiments using Binary IMGO (BIMGO), including comparisons with state-of-the-art binary metaheuristic algorithms. Ablation experiments investigating the individual contributions of the male competition, ownership, and mating strategies. Statistical analyses, including Friedman and Wilcoxon signed-rank tests, for evaluating the significance and robustness of the obtained results. Experimental results, convergence data, and performance measures related to classification accuracy, feature reduction, and optimization performance. The repository is provided to support reproducibility, transparency, and further research on chaos-enhanced metaheuristic optimization, binary optimization, and metaheuristic-based feature selection.