Predicting Fifth Generation (5G) Network Coverage using Multilayer Perceptron Neural Network
Abstract
The fundamental step needed for planning and optimizing any wireless network during its early phase of deployment involves the estimation of radio coverage. Fifth generation (5G) telecommunication systems in Nigeria are in their early stages; hence, there is a need to develop effective mobile coverage prediction models with high accuracy and minimal complexity. In this study, measurements of 5G signal strength on 3.5GHz frequency operation were carried out at three different locations in Rivers State, Niger Delta region of Nigeria. Atmospheric variables at the study locations were collected with a Power Data Access Viewer (DAV) Web Mapping Application. The path loss prediction model developed in this study is a function of technical variables (distance between transmitter and receiver as well as antenna heights of the transmitter and receiver) and atmospheric variables (air temperature, wind speed, wind direction, and precipitation). A multilayer perceptron (MLP) neural network was employed to develop the proposed model. The MLP neural network model prediction performance was based on the mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (r2). The model’s validity was assessed by comparing its results with those of empirical path loss models. The MLP neural network achieved an R-square score of 0.88, indicating that it explained 88% of the variability in the dataset. The MLP model demonstrated a substantial improvement in accuracy, reducing the RMSE to 3.80 dB compared with the standard 8 dB benchmark for tuned models. The results obtained by the MLP neural network model suggest that atmospheric conditions play a significant role in the evaluation of 5G mobile signal analysis.