Evaluation of Characteristics for NN and CNN in Path Loss Prediction

In path loss prediction, recently machine learning approaches such as the convolutional neural network (CNN) have been attracting interest. Various models have been proposed that estimate the path loss characteristics by automatically extracting parameters from satellite photographs of buildings and so on. However, at present neither the estimation tendency nor the characteristics of the path loss estimation using NN and CNN are sufficiently clear. In this paper, we propose a CNN model constructed using the same concept as the NN model and evaluate the differences and tendencies of the RMS error.