论文部分内容阅读
提出改进的预报相对误差法选择岭回归参数k,以平均预报相对误差替代预报残差平方和,并抑制过拟合。该方法应用于苯乙酰胺类除草农药定量构效关系的二次建模,效果良好,预报精度优于残差平方和方法。
An improved relative error method is proposed to select the ridge regression parameter k, and the relative error of the average prediction is used to replace the sum of squared residuals of the prediction residuals, and the over-fitting is suppressed. The method is applied to the secondary modeling of the quantitative structure-activity relationship of phenylacetamide herbicidal pesticides with good effect and better prediction accuracy than residual square sum method.