Reliable comparison for power amplifiers nonlinear behavioral modeling based on regression trees and random forest.

This work evaluates the construction of feature extraction nonlinear behavioral models based on Regression Trees and Random Forest techniques. A framework to evaluate the effectiveness with enough-accuracy regressor models are evaluated to aid in the design of a digital predistorter (DPD) for the power amplifier (PA) linearization. The comparison with a conventional memory polynomial model (MPM) and two ensemble learning models is performed to reveal the ability in decision and region identification without overfitting for the Regression Tree and a Random Forest algorithms.