Optimizing Twin Support Vector Machine for Breast Cancer Classification Using a Hybrid Metaheuristic Algorithm Based on Multi-Agent Reinforcement Learning

Document Type : Research Paper

Authors
Omid SolaymaniFard
Ayda Rahimi
Ferdowsi University of Mashhad
10.22034/mmr.2026.11745
Abstract
Breast cancer diagnosis requires accurate classification methods to support early detection and improve clinical outcomes. This study aims to propose a hybrid optimization framework to enhance Twin Support Vector Machine (‎Twin SVM‎) performance for breast cancer classification. The framework integrates four marine-inspired metaheuristic algorithms — Particle Swarm Optimization (PSO), Whale Optimization Algorithm (WOA), Salp Swarm Algorithm (SSA), and Marine Predators Algorithm (MPA) — as cooperative structure. Q-Learning dynamically controls the exploration-exploitation balance by adaptively selecting the most effective search operator at each iteration, preventing premature convergence and improving global search capability. Experimental results on two benchmark breast cancer datasets demonstrate that the proposed method achieves superior accuracy, ‎precision‎, and recall compared to conventional Twin SVM and single-algorithm optimizers.
Keywords
Twin Support Vector Machine‎
Breast Cancer Classification‎
Hybrid Metaheuristic ‎Algorithm‎‎
Multi Agent Reinforcement