基于最小二乘支持向量机的农田水汽通量建模

Modeling Surface Water Vapor Flux Exchange over a Summer Maize Field Based on Least Squares Support Vector Machines

支持向量机是在统计学习理论基础上发展起来的一种新型的机器学习方法,已在模式识别、非线性建模等领域中得到了应用．本文将最小二乘支持向量机方法应用于农田水汽通量的建模中,并同前馈反向传播神经网络的建模性能进行了比较．结果表明,最小二乘支持向量机方法具有可调参数少、学习速度较快等优点,具有更好的推广能力,以更高的精度建立农田水汽通量模型．模型的敏感性分析进一步显示,用最小二乘支持向量机方法建立的农田水汽通量模型是合理可行的．

Support vector machines （SVM） are a kind of novel machine learning methods based on statistical learning theory, which have been developed for solving pattern recognition and nonlinear modeling problems. This paper applies Least Squares Support Vector Machines （LS-SVM） to model farmland vapor flux, and compares the modeling results with BP network. The study shows that LS-SVM has only two parameters and it can be trained fastly. The LS-SVM has good ability of modeling nonlinear process and good generalization. The farmland vapor flux modeling has high accuracy. The sensitivity analysis of the derived LS-SVM also shows the vapor flux model is reliable.