基于多种机器学习及其堆叠式集成方法的月尺度北极海冰预测研究

岳瀚栋, 窦挺峰, 李润奎, 丁明虎, 效存德

Monthly-scale Arctic sea ice extent prediction based on multiple machine learning and stacking ensemble methods

Handong YUE, Tingfeng DOU, Runkui LI, Minghu DING, Cunde XIAO

表4 各模型给出的2000年海冰范围预测精度 (×106 km2)

Table 4 Prediction accuracies of sea ice extent in 2000 of various models

模型名 RMSE MAE LR 1.50 1.35 SVR 1.38 1.23 LGB 0.37 0.31 XGB 0.39 0.33 CAT 0.36 0.30 DF 0.52 0.46 Ens 0.35 0.29 DNN 1.22 1.03 CNN 1.05 0.88 ConvLSTM 0.80 0.77