Document Type : Research Paper
Abstract
This study presents a comparative analysis of signal propagation models using a Radial Basis Function Neural Network in Effurun, Delta State, Nigeria. Conventional empirical models such as Okumura Hata, ECC 33, and SUI often lose accuracy when applied to fast growing urban areas with complex terrain, building density, and tropical weather. Field measurements of received signal strength were collected across eight locations through a six month drive test. These data formed the basis for training, validation, and testing of the neural network developed in MATLAB Simulink. The Radial Basis Function Neural Network learned the nonlinear relationship between distance, environmental factors, and signal strength, then predicted path loss for the study area. Model performance was evaluated using mean square error, root mean square error, standard deviation, and coefficient of efficiency. Results show that the neural network outperformed all empirical models across the eight routes. The best performance recorded a root mean square error of2.15 dB and a coefficient of efficiency close to unity. The findings show that a data driven approach captures local propagation behavior more accurately than generic empirical formulations. The study provides a reliable prediction tool for wireless network planning in Nigerian urban environments and offers a framework adaptable to other tropical cities with similar characteristics.
