Coverage Prediction and REM Construction for 5G Networks in Band n78

Abstract

Currently, a great number of commercial fifth-generation (5G) networks are deployed on the mid-band, especially in the range between 3.3 GHz and 3.8 GHz, denominated Band n78. Therefore, a radio environment map (REM) construction is a meaningful task for a network operator to indicate the service areas of 5G cellular systems, improve network planning, and handle mobility. Thus, we propose a novel approach to predict the coverage of outdoor-to-indoor propagation for 5G mid-band operational networks, based on the extremely randomized trees regressor (ERTR) algorithm. Then, we construct a REM to improve the visualization of the results and easily detect coverage holes and traffic hotspots.For this purpose, we utilize a dataset of channel measurements in a building of Sapienza University of Rome, Italy. Furthermore, we use three error metrics: relative error, mean absolute error (MAE), and root mean square error (RMSE) to validate our proposed ERTR-based scheme. For comparison purposes, we evaluate the performance of five additional machine learning (ML) regression algorithms. Satisfactorily, the proposed ERTR technique outperforms the comparative approaches by improving the accuracy of coverage prediction in all scenarios.