基于多源环境变量和随机森林模型的江西省耕地土壤pH值空间预测

钟骁勇, 李洪义, 郭冬艳, 谢模典, 赵婉如, 胡碧峰

Spatial distribution prediction of soil pH in arable land of Jiangxi Province based on multi-source environmental variables and the random forest model

ZHONG Xiaoyong, LI Hongyi, GUO Dongyan, XIE Modian, ZHAO Wanru, HU Bifeng

表3 使用不同训练集时RF和OK模型预测精度比较

Tab.3 Comparison of model performance with different training and validation dataset ratios

预测模型 抽样比 训练集 验证集 r ME MAE RMSE r ME MAE RMSE RF-A 8∶2 0.970 -0.004 0.121 0.167 0.599 -0.009 0.291 0.392 7∶3 0.970 -0.004 0.121 0.167 0.567 -0.014 0.299 0.401 6∶4 0.970 -0.004 0.122 0.168 0.566 -0.018 0.302 0.406 RF-B 8∶2 0.966 -0.005 0.127 0.175 0.548 -0.023 0.307 0.413 7∶3 0.967 -0.005 0.127 0.175 0.525 -0.019 0.308 0.418 6∶4 0.967 -0.005 0.127 0.174 0.504 -0.014 0.314 0.425 OK 8∶2 0.679 0.002 0.261 0.364 0.559 0.009 0.311 0.409 7∶3 0.683 0.001 0.259 0.357 0.588 0.007 0.289 0.394 6∶4 0.652 0.002 0.278 0.379 0.541 0.010 0.328 0.414