Machine Learning-Based Optimization Algorithms for Spam SMS Classification

Modern communication networks are plagued by spam Short Message Service (SMS), which invade users' personal spaces and pose significant security threats. This study proposes a novel spam SMS categorization method that combines Machine Learning (ML) techniques with the Golden Search Optimizer (GSO) and the Whale Optimization Algorithm (WOA). The hybrid Golden Search Optimizer–Whale Optimization Algorithm (GSOWOA) optimization technique is integrated with Random Forest (RF) models to improve performance and convergence time. Experimental findings reveal that categorization accuracy is much higher than that of classical ML approaches. By fine-tuning model hyperparameters, the optimization process reduces both false positive and false negative classifications, enhancing overall performance. The results demonstrate an effective and systematic approach for improving spam SMS categorization systems. This strategy has been beneficial for optimally tuning hyperparameters, leading to high classification performance for models such as Extreme Gradient Boosting (XGBoost) and RF. The main experimental evaluation indicates that the proposed framework achieved a total accuracy of 98.9% and a precision of 97.85%, with a 35% faster convergence rate than conventional protocols and general metaheuristic methods. Notably, GSOWOA demonstrated a strong resistance against overfitting, coupled with computational efficiency, making it viable for real-time spam detection on edge devices. These results provide evidence of the practical benefits of implementing a hybrid optimization process to achieve high performance and optimal resource utilization for SMS filtering.