Enhancing 4G/LTE Network Path Loss Prediction with PSO-GWO Hybrid Approach

Abstract

In a 4G/LTE network, path loss models are crucial for efficient planning, interference estimation, frequency allocation, and optimization of cell parameters, all essential for the network's overall performance and planning. The Hata, COST231, Egli, and SUI models are among the most widely used in urban and suburban environments. However, these models rely on fixed parameters, which can cause inaccuracies when applied to measured data. This paper uses data collected from the north-central region of Algeria to calculate the path loss for 4G LTE (1800 MHz). The measured path loss is compared with the empirical models mentioned above. The best model for estimating the measured path loss is then optimized using three well-known evolutionary algorithms: Particle Swarm Optimization (PSO), Grey Wolf Optimizer (GWO), and Covariance Matrix Adaptation Evolution Strategy (CMA-ES). Furthermore, a hybrid optimization model, PSO-GWO, is proposed to improve prediction accuracy. The performance of each optimization method is evaluated using various error metrics, including Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Standard Deviation (STD), Mean Error (ME), and Mean Absolute Percentage Error (MAPE). Results demonstrate that the hybrid techniques outperform the original Hata and COST231-Hata models, with the PSO-GWO hybrid yielding the lowest RMSE, achieving a significant reduction compared to the traditional models.