SIGNAL POWER LOSS PREDICTION USING ARTIFICIAL INTELLIGENCE

Resumo

In this study, we conducted a machine learning approach to propose a mobile telephony signal propa-
gation model using regression. The acquisition of signal propagation models provides relevant indicators about

the network signal quality offered to users. Although the most advanced technology is 5G, a study was developed
using 3G and 4G data, since the implementation of 5G in Brazil is still a distant scenario. Signal power loss
data from a single operator were collected through the G-Net Track application and processed using Haversine.
This enabled the application of linear regression technique to obtain a model representing signal power loss. The
regression-generated result was compared to nine literature models, including Rappaport, Okumura-Hata, ECC33,
Modified Cost231-Hata, SUI, Extended SUI, Walfish-Ikegami, and Ericson999. The results from the sampled data
indicate that the literature models do not adequately represent the signal behavior, and that the application of linear

regression produces a solution capable of representing, in a more realistic manner, the behavior of 3G and 4G sig-
nal power loss concerning the transmitting antenna. Through this study, we aim to comprehend the heterogeneity

of the network infrastructure and contribute by providing research that aids in the formulation of public policies to
enhance the Brazilian telecommunications system.