MDSH-SVM for High-Dimensional Support Vector Machine Optimization

Support Vector Machines (SVMs) excel in classification but require careful hyperparameter tuning and feature selection to handle high-dimensional data. We propose MDSH-SVM, a novel hybrid algorithm integrating Hunger Games Search (HGS) and Slime Mould Algorithm (SMA) for simultaneous feature selection and SVM parameter optimization. Extensive experiments on publicly available high-dimensional datasets demonstrate that MDSH-SVM achieves statistically significant improvements over GA-SVM, PSO-SVM, and other state-of-the-art heuristics in accuracy, model compactness, and runtime efficiency. Detailed sensitivity analyses and ablation studies further validate the robustness and adaptability of the proposed method across varying data characteristics.