以程海为例用支持向量机回归算法预测叶绿素a浓度

Using Support Vector Regression Algorithm to Predict Chlorophyll-a Concentrations with Chenghai Lake for Example

应用支持向量机回归(SVR)算法预测程海富营养化水体叶绿素a(Chl-a)的浓度,用留一法交叉验证(LOOCV)优化SVR预测模型的参数,并根据平均相对误差(MRE),讨论SVR预测模型的准确性。结果表明:用径向基核函数构建的SVR预测模型预测结果最优;SVR预测模型的预测值和实测值具有很好的一致性,相关系数为0.938,MRE为12.30%。SVR预测模型的建模结果优于人工神经网络(BP-ANN)预测模型,说明SVR算法能够准确预测Chl-a浓度。

The support vector regression （SVR） algorithm was used to predict the concentration of chlorophyll-a （Chl-a） of eutrophication water in Chenghai Lake, and the leave-one-out cross-validation （LOOCV） method was used to optimize the model parameters. Then the prediction accuracy of SVR model was discussed on the basis of the mean relative error （MRE）. The results demonstrated that the SVR model built by radial basis kernel function （RBF） had the optimal predictive ability. The predicted values of SVR were in good consistency with the measured values of experiment. The correlation coefficient （R） and MRE of SVR model could reach 0.938 and 12.30% , respectively. It was found that the modeling results of SVR were better than that of back propagation artificial neural networks （BP-ANN） , suggesting that SVR was a valuable tool for the prediction of Chl-a.