基于偏最小二乘回归的水稻腾发量建模

Modeling Rice Evapotranspiration with Partial Least-Squares Regression

利用气象因子计算水稻腾发量过程中 ,各自变量之间经常存在多重相关性 ,从而导致传统的多元回归模型 (基于最小二乘法 )的失真 ,丧失稳健性 ,预测精度降低。该文采用偏最小二乘回归建模 ,借助主成分分析与典型相关分析的思路 ,采用成分提取的方法 ,有效地解决了各气象因子之间的多重相关性 ,建立了水稻腾发量模型 ,并对模型进行了辅助分析 。

Calculating the evapotranspiration with weather data, it was found that some independent variables had interactions with each other. This phenomenon can distort and destabilize the multivariate regression model of traditional least square method. The partial least-square regression for model was applied to model the rice evapatranspiration base on the main component analysis and typical correlation analysis of data. The model of rice evapotranspiration was suggested to solve the teractive correlation among the independent variables (weather parameters.). The model was found to be able to give satisfactory predictions.