| Abstract |
This study presents a machine learning-based approach to forecast Allocative Localization Error (ALE) in Wireless Sensor Networks (WSNs), addressing challenges such as dynamic network topologies and resource constraints. The approach utilizes Radial Basis Function (RBF) models enhanced with advanced optimization algorithms, including Coot Optimization Algorithm (COA), Smell Agent Optimization (SAO), and Northern Goshawk Optimization (NGO) to improve ALE prediction accuracy. Hybrid models (RBCO, RBSO, and RFNG) are developed by integrating these optimization techniques, which refine critical RBF parameters, such as spread and center selection, through iterative optimization. Furthermore, an ensemble framework (RSNC) combines all three optimizers with RBF to achieve superior performance. The proposed methods are validated using R2 and RMSE metrics, demonstrating their ability to minimize ALE, optimize resource allocation, and extend network lifespans. The study highlights the practical applicability of these models in real-world scenarios, such as environmental monitoring and industrial automation, offering enhanced efficiency and economic benefits. The RFNG model, in particular, achieved the lowest Mean Absolute Relative Error (MARE) of 0.049, demonstrating superior performance compared to other approaches in the test section. Moreover, RBNG obtained 0.069 and 0.978 values for the RMSE and R2, respectively, which were the most suitable values compared to other models, namely RBGO, RBSO, RSNC, and RBF. The results indicate that the proposed hybrid models significantly improve the prediction of ALE, leading to more efficient node deployment and better network management. This research provides valuable insights into leveraging machine learning for WSN optimization, benefiting researchers, network engineers, and industries relying on sensor networks for applications such as environmental monitoring, smart cities, and asset tracking. © 2025 |
