A Higher Order Prediction Model of Populus Simonii’s Net Photosynthetic Rate Based on Improved Gradient Boosting Method
A Higher Order Prediction Model of Populus Simonii’s Net Photosynthetic Rate Based on Improved Gradient Boosting Method
复制标题
DOI：
10.1007/978-981-33-6378-6_28
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Zhang Xiaoyu;Ji Xinyue-;Song Yuepeng-;Zhang Deqiang-;Qingshu Fang
中科院分区：
文献类型：
--
作者：
Zhang Xiaoyu;Ji Xinyue-;Song Yuepeng-;Zhang Deqiang-;Qingshu Fang
本研究以548个杨树个体为核心种群，建立了杨树单株净光合速率预测模型。simonii。我们测量了这些个体的光合特性数据（净光合速率、气孔导度、细胞间CO2浓度、水分利用效率）和叶片表型数据（叶面积、长、宽、周长、长宽比、叶形因子）。我们首先利用平均连锁聚类和PAM聚类将这些个体分为三个亚群。然后，我们使用不同的机器学习方法来预测基于叶片表型数据的净光合速率。特别是在梯度增强方法中，提出了收缩估计准则和迭代停止准则来增强模型。交叉验证结果表明，我们的模型具有较高的预测准确率（三个亚群分别为90.87%、88.34%和89.26%），同时改善了其他机器学习方法的过拟合。
We develop a net photosynthetic rate prediction model ofPopulus simoniiby selecting 548 individuals as core populations that represent almost the entire geographic distribution ofP. simonii. We measure photosynthetic characteristic data (net photosynthetic rate, stomatal conductance, intercellular CO2 concentration, water use efficiency) and leaf phenotypic data (leaf area, length, width, perimeter, length-width ratio, leaf shape factor) of these individuals. We first classify these individuals into three subpopulations by utilizing average linkage clustering and PAM clustering. Then we use different machine learning methods to predict net photosynthetic rate based on leaf phenotypic data. Especially in gradient boosting method, the criterion of shrinkage estimator and the iteration-stopping criterion are put forward to enhance the model. The cross-validated results show that our model has high prediction accuracy (90.87%, 88.34%, and 89.26%, respectively in three subpopulations) and improve overfitting in other machine learning method at the same time.