Performance analysis of artificial neural networks for predicting propagation losses in suburban environments for 4G LTE and 5G networks
This study analyzes two distinct approaches for predicting path loss at frequencies of 800 MHz, 1800 MHz, and 2600 MHz in suburban areas. These frequencies are commonly utilized in broadcasting, 4G LTE, and 5G networks. Two models of Artificial Neural Networks (ANN) were implemented: an Error-Based...
| Autores: | , , |
|---|---|
| Formato: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2025 |
| País: | Brasil |
| Recursos: | Instituto Federal de Educação, Ciência e Tecnologia da Paraíba (IFPB) |
| Repositorio: | Revista Principia |
| Idioma: | inglés |
| OAI Identifier: | oai:ojs.periodicos.ifpb.edu.br:article/8532 |
| Acesso em linha: | https://periodicos.ifpb.edu.br/index.php/principia/article/view/8532 |
| Access Level: | acceso abierto |
| Palavra-chave: | 4G LTE 5G networks artificial neural networks path loss |
| Resumo: | This study analyzes two distinct approaches for predicting path loss at frequencies of 800 MHz, 1800 MHz, and 2600 MHz in suburban areas. These frequencies are commonly utilized in broadcasting, 4G LTE, and 5G networks. Two models of Artificial Neural Networks (ANN) were implemented: an Error-Based Neural Network (EBNN), which incorporates error correction by combining empirical propagation models with an ANN, and a Terrain Parameters Based Neural Network (TBNN), which uses input parameters commonly applied in related studies, such as the distance from the transmitter to the receiver, receiver altitude, average terrain level, and azimuth angle between the transmitter and receiver. The performance of these models was evaluated using root mean square error (RMSE) and the Wilcoxon rank-sum test, comparing them with empirical propagation models such as SUI, ECC-33, Ericsson, and TR 25.942. The results were then compared with data obtained from a measurement campaign conducted along three routes in the city of Natal, Brazil. The findings from both simulations and actual measurements showed good metric alignment, particularly highlighting the performance of the error-based model. The primary contribution of this study is demonstrating that these techniques enable more accurate prediction of signal information, thereby reducing errors in the planning and implementation of wireless networks. |
|---|