A Random Forest Algorithm Combined with Bayesian Optimization for Atmospheric Duct Estimation

Inversion of atmospheric ducts is great importance in the field performance evaluation for radar and communication systems. Since model parameters machine learning play a crucial role prediction performance, this paper develops random forest (RF) integrated with Bayesian optimization (BO) called BO-RF duct prediction, BO adopted to determine appropriate during training process. In addition, K-fold cross-validation (CV) method also incorporated into obtain best partition overcome overfitting problem. To test proposed model, results obtained by are compared other commonly used methods, such as classical RF, extreme gradient boosting (XGBoost) with/without BO, K-nearest neighbor (KNN) BO. Comparisons demonstrate that has accuracy anti-noise ability estimation parameters.