Path-Loss Prediction of Millimeter-wave using Machine Learning Techniques

Abstract

Millimeter-wave communication systems design require accurate path-loss prediction, critical to determining coverage area and system capacity. In this work, four machine learning algorithms are proposed for path-loss prediction in an indoor environment for 5G millimeter-wave frequencies, from 26.5 to 40 GHz. They are artificial neural network, support vector regression, random forest, and gradient tree boosting. We compare their performances, including extensions of the empirical path-loss models alpha-beta-gamma and close-in frequency-dependent exponent incorporating the number of crossed walls. The results show that the ML techniques improve the prediction accuracy of empirical models.