An Evaporation Duct Feature Importance Discrimination Method Based on Random Forest Algorithm

The evaporation duct exerts a significant influence on electromagnetic wave propagation. Given its high occurrence probability in the East China Sea and South China Sea regions of China, the effective prediction and timely warning of maritime evaporation ducts hold substantial importance for the nation. This study focuses on the geographical area spanning 100°E to 130°E longitude and 0°N to 45°N latitude. Using meteorological data from the Weather Research and Forecasting (WRF) model for this region, features were extracted. The random forest algorithm was then employed to train a model for predicting the presence or absence of an evaporation duct at locations within the area. The results indicate that the prediction accuracy and computational efficiency of the proposed model for the selected region are superior to those achieved using the WRF model alone. By analyzing feature importance, a ranking of the significance of features for predicting evaporation duct occurrence at each location was obtained, further enhancing computational efficiency.