Metaheuristic Optimization Algorithms for the Control and Management of Arsenic Concentration in the Eastern Province of Saudi Arabia

Accurate and reliable prediction models to monitor and manage arsenic concentration in groundwater to ensure safe drinking water and protect public health in arsenic-affected regions are paramount. This study evaluates the performance of three models: Artificial Neural Network (ANN), ANN optimized with Chicken Swarm Optimization (ANN-CSO), and ANN optimized with Whale Optimization Algorithm (ANN-WOA) for predicting arsenic (As) concentration in groundwater in the Eastern Province of Saudi Arabia. The ANN-WOA model demonstrated the best performance during the training phase with the lowest RMSE (0.1195), MSE (0.0143), MAE (0.0909), and the highest R² (0.9627) and PCC (0.9813). In the testing phase, despite a performance decline, ANN-WOA maintained superior accuracy with an R² of 0.4736, compared to ANN (0.3012) and ANN-CSO (0.2915). The Mean Absolute Error (MAE) analysis indicated that ANN had the lowest MAE (0.4079), while ANNWOA had a slightly higher MAE (0.4194) but better overall metrics. Violin plots showed wider distributions for predicted values, with ANN-WOA displaying the highest variability. The study highlights the environmental significance of accurate prediction models for effective monitoring, timely interventions, and efficient resource management to reduce arsenic exposure risks. Although ANN-WOA is the most robust model, further refinement is needed to enhance reliability and reduce overestimation, ensuring safe drinking water and protecting public health in arsenic-affected regions. Accurate prediction models like ANN-WOA support sustainable groundwater management and compliance with safety standards.