Explainable Deep-Learning-Based Path Loss Prediction from Path Profiles in Urban Environments

This paper applies a deep learning approach to model the mechanism of path loss based on profile in urban propagation environments for 5G cellular communication systems. The proposed method combines log-distance line-of-sight scenarios and deep-learning-based non-line-of-sight cases. Simulation results show that outperforms conventional models when operating 3.5 GHz frequency band. standard deviation prediction error was reduced by 34% compared models. To explain internal behavior model, which is black box nature, eight relevant features were selected linear regression approach. accuracy explanatory reached 72% it used model. Furthermore, also evaluated non-standalone New Radio network environment Taipei City. real-world measurements can be 30–43% In addition, transparency 63% realistic network.