基于梯度提升树的土壤速效磷高光谱回归预测方法 
下载： 986次
Predicting Soil Available Phosphorus by Hyperspectral Regression Method Based on Gradient Boosting Decision Tree
安徽农业大学信息与计算机学院, 安徽 合肥 230036
图 & 表
图 1. Stacking方法
Fig. 1. Stacking method
下载图片 查看原文
图 2. 室内高光谱采集系统
Fig. 2. Indoor hyperspectral acquisition system
下载图片 查看原文
图 3. 土壤高光谱反射率。(a)原始光谱;(b)平滑后光谱
Fig. 3. Hyperspectral reflectance of soil. (a) Original spectra; (b) smoothing spectra
下载图片 查看原文
图 4. 线性和非线性PLS中不同LV对应的方均根误差值
Fig. 4. fRMSE values of different LV numbers in linear and nonlinear PLS
下载图片 查看原文
图 5. 梯度提升树模型参数优化。(a) Rloss=Fls,Rn_estimators=100;(b) Rloss=Fhuber,Rn_estimators=200; (c) Rloss=Fquantile,Rn_estimators=200;(d) Rloss=Flad,Rn_estimators=310
Fig. 5. Parameter optimization of GBDT model. (a) Rloss=Fls, Rn_estimators=100; (b) Rloss=Fhuber, Rn_estimators=200; (c) Rloss=Fquantile, Rn_estimators=200; (d) Rloss=Flad, Rn_estimators=310
下载图片 查看原文
图 6. 不同模型集成算法的结果。 (a)基于建模集的随机森林结果; (b)基于测试集的随机森林结果 ;(c)基于建模集的提升树结果; (d)基于测试集的提升树结果 ;(e)基于建模集的梯度提升树结果 ;(f)基于测试集的梯度提升树结果
Fig. 6. Results of different model integration algorithms. (a) Results of random forest based on modeling set; (b) results of random forest based on testing set; (c) results of boosting tree based on modeling set; (d) results of boosting tree based on testing set; (e) results of GBDT based on modeling set; (f) results of GBDT based on testing set
下载图片 查看原文
表 1土壤速效磷含量的统计参数
Table1. Statistical parameters of soil available phosphorus content
| Type | Sample | Max /(mg·kg-1) | Min /(mg·kg-1) | Average /(mg·kg-1) | Standard deviation /(mg·kg-1) |
|---|
| Total | 193 | 34.96 | 0.03 | 10.56 | 9.36 | | Training | 144 | 34.96 | 0.03 | 10.94 | 9.49 | | Testing | 49 | 32.24 | 0.60 | 9.01 | 8.99 |
|
查看原文
表 2最优单模型的测试结果
Table2. Testing results of optimal single model
| Modeling method | Training set | Testing set | Prediction level(testing set) | Parameter |
|---|
| fRPD | | | R2 | fRPD | R2 |
|---|
| PLS | 1.66 | 0.73 | 1.65 | 0.68 | B | RLVs=11 | | RBF-PLS | 1.58 | 0.71 | 1.79 | 0.73 | B | RLVs=11, Rgamma =0.016 | | Sigmoid-PLS | 1.55 | 0.70 | 1.75 | 0.73 | B | RLVs =10, =0.00085,Rcofe0cofe0=4.5 | | SVR | 1.60 | 0.74 | 1.53 | 0.69 | B | C=10000 | | RBF-SVR | 1.70 | 0.76 | 1.66 | 0.72 | B | C=2000000, Rgamma =0.0028 | | Sigmoid-SVR | 1.59 | 0.73 | 1.55 | 0.70 | B | C=1011, Rgamma =0.000001,Rcofe0=0 | | Ridge | 1.60 | 0.74 | 1.50 | 0.69 | B | RAlpha=0.001 | | RBF-Ridge | 1.55 | 0.74 | 1.50 | 0.70 | B | RAlpha=0.00006, Rgamma =0.01 | | Sigmoid-Ridge | 1.52 | 0.72 | 1.50 | 0.69 | B | RAlpha=4×10-7,Rgamma =0.0005, Rcofe0=0.9 |
|
查看原文
表 3多种模型组合结果
Table3. Results of multi-model combination
| Ensemble method | Training set | Testing set | Prediction level(testing set) | Parameter |
|---|
| fRPD | | | R2 | fRPD | R2 |
|---|
| Random forest | 2.10 | 0.84 | 2.08 | 0.84 | A | Rn_estimators, Rmax_depth=5 | | Boosting tree | 2.86 | 0.90 | 2.12 | 0.82 | A | Rn_estimators =300, Rlearning_rate=0.01,Rmax_depth=5, Rloss=Flinear | | GBDT | 2.56 | 0.88 | 2.55 | 0.86 | A | Rn_estimators =310, Rlearning_rate=0.29,Rmax_depth=4, Rloss=Flad |
|
查看原文
金秀, 朱先志, 李绍稳, 王文才, 齐海军. 基于梯度提升树的土壤速效磷高光谱回归预测方法[J]. 激光与光电子学进展, 2019, 56(13): 131102. Xiu Jin, Xianzhi Zhu, Shaowen Li, Wencai Wang, Haijun Qi. Predicting Soil Available Phosphorus by Hyperspectral Regression Method Based on Gradient Boosting Decision Tree[J]. Laser & Optoelectronics Progress, 2019, 56(13): 131102.
PDF全文
