Algorithm 1 Ensemble Feature Selection with SHAP Refinement

Input: Dataset D with features X, labels y, threshold $\theta$ Normalize X and split into train/test sets function Fitness(solution)       Convert $solution$ to binary mask       if no feature selected then             return $1.0$       end if       Train SVM on selected features       return $1-$ accuracy end function Define a binary optimization problem with Fitness as objective Run HHO, mGTO, and ZOA to generate feature masks Apply majority voting to obtain ensemble mask Train SVM using features selected by ensemble mask Compute SHAP values; rank features by mean absolute value Select top features achieving $\theta$ cumulative SHAP contribution Output: Optimized feature subset $X_{\mathrm{SHAP}}$