Machine Learning Based Mm-wave 60 GHz Channel Modeling for 5G Wireless Communication Systems

Abstract

In this paper, based on mm-wave 60 GHz channel measurements performed in corridor and large hall for both LOS and NLOS scenarios, channel statistical parameters are investigated based on machine learning (ML) methods. Specifically, path loss and delay spread are predicted by using back propagation (BP), support vector machine (SVM) and genetic algorithm (GA) neutral network models. Results show that the GA+ SVM model can fit measurement data excellently. As the proposed GA+ SVM model better approaches measurement data in the sense of signal correlation coefficients are larger and errors are smaller than the other models. More importantly, results show the advances of ML in channel modeling, or the expensive of channel measurements can be replaced as the ML methods can accurately predict channel parameters. The presented results are useful in design of 5G wireless communication systems and system development.