基于SHAP可解释特征优选的乌克兰作物制图研究OA

The precise and efficient delineation of distribution of winter wheat,corn,and soybeans is of vital importance for agricultural resource management,crop planting planning,and agricultural policy formulation.Focused on the agricultural regions of Ukraine,based on Sentinel-2 remote sensing images,we combined the feature selection method of Spearman correlation coefficient with SHAP,to optimize the mapping process of major crops such as winter wheat,corn,and soybeans.We explored the effectiveness of seven combination strategies with three machine learning models in distinguishing major crop types.The results showed that①the feature selection method combining Spearman correlation coefficient with SHAP demonstrated the best performance.②Based on the optimal features derived from Spearman correlation coefficient and SHAP,the LightGBM algorithm achieved the highest classification accuracy with overall accuracy and Kappa coefficient of 95.42%and 90.83%for winter wheat mapping,and overall accuracy and Kappa coefficient of 96.19%and 94.28%for corn,soybeans mapping,respectively.The classification results demonstrated high consistency with statistical yearbook data,achieving area extraction accuracy of 97.31%for winter wheat,97.55%for soybean,and 99.61%for corn.