The paper introduces a novel feature selection method called S-shaped Grey Wolf Optimizer-based FOX (FOX-GWO), which combines the FOX algorithm with Grey Wolf Optimizer (GWO) and incorporates an S-shaped transfer function. The FOX-GWO algorithm aims to address the limitations of the basic FOX algorithm, particularly in high-dimensional problems, by enhancing local search capabilities and improving binary conversion. The S-shaped transfer function facilitates the transformation of continuous values into binary form, making it suitable for feature selection tasks. The proposed method was evaluated on 18 datasets and compared with five other state-of-the-art optimizers. The results show that FOX-GWO outperforms other methods in terms of average accuracy, reduced feature dimensionality, and average fitness value. The algorithm demonstrates superior performance in 83.33% of the datasets for average accuracy, 61.11% for reduced feature dimensionality, and 72.22% for average fitness value. The study highlights the practical value and potential of FOX-GWO in advancing feature selection in complex data analysis, enhancing model prediction accuracy.The paper introduces a novel feature selection method called S-shaped Grey Wolf Optimizer-based FOX (FOX-GWO), which combines the FOX algorithm with Grey Wolf Optimizer (GWO) and incorporates an S-shaped transfer function. The FOX-GWO algorithm aims to address the limitations of the basic FOX algorithm, particularly in high-dimensional problems, by enhancing local search capabilities and improving binary conversion. The S-shaped transfer function facilitates the transformation of continuous values into binary form, making it suitable for feature selection tasks. The proposed method was evaluated on 18 datasets and compared with five other state-of-the-art optimizers. The results show that FOX-GWO outperforms other methods in terms of average accuracy, reduced feature dimensionality, and average fitness value. The algorithm demonstrates superior performance in 83.33% of the datasets for average accuracy, 61.11% for reduced feature dimensionality, and 72.22% for average fitness value. The study highlights the practical value and potential of FOX-GWO in advancing feature selection in complex data analysis, enhancing model prediction accuracy.