Radio Environment Map Construction Based on Random Forest Regression

Abstract

Radio environment map is an effective tool for electromagnetic environment awareness and spectrum resource management, which can be constructed by spatial interpolation methods. In order to improve the quality of radio environment map construction, this paper applies machine learning technologies to spatial interpolation and proposes a random forest spatial interpolation algorithm improved by the golden jackal optimizer. Initially, establish a random forest regression model for spatial interpolation based on machine learning method by using data of measurement points. Then optimize parameters of the model with the help of golden jackal optimizer. At last, utilize the optimized random forest regression model to interpolate missing data of the radio environment map. Experiment results demonstrate that the proposed algorithm has better performance in construction accuracy compared to classical spatial interpolation algorithms.