Prediction of urban propagation loss using regression trees

Radio path loss prediction is crucial. For instance it permits to determine best station coverage or reduction of interference. This paper presents a recent statistical method called CART (classification and regression trees) applied to the prediction of the propagation path loss. This non-parametric regression method is based on binary decision trees and considers non-linear relationships between parameters. The adaptation of this technique has been tested on measured data in urban area. This method has also been compared with linear models