Development of a Hybrid Machine Learning Path Loss Model for Cellular Networks in Maritime Environments Using Regression-Based Fusion

Abstract

 Accurate path loss modelling is critical for optimizing LTE network coverage and performance, particularly in maritime environments where conventional models often struggle to provide precise estimations due to dynamic environmental conditions. This study presents a hybrid path loss model that integrates machine learning (ML) with traditional empirical models using a regression-based fusion technique. The proposed approach enhances prediction accuracy by dynamically adjusting model coefficients based on real-time environmental factors. A case study of the Escravos water in Delta State, Nigeria, demonstrates the model's effectiveness compared to standard empirical models such as COST-231 and Okumura-Hata. The results indicate significant improvements in prediction accuracy, with a reduced mean squared error (MSE) and enhanced adaptability to environmental variations. The findings suggest that the proposed decision tree particle swarm optimization (DT-PSO-COST231) path loss model can serve as a valuable tool for LTE network planning in maritime environments.