论文部分内容阅读
将数值遗传算法同非线性PLS结合,改进和完善了非线性PLS,推导了指数、对数、倒数和Sigmoid函数的公式,构造了可以处理多种非线性函数关系的算法,可用于解决复杂的结构与性能相互关系.用该法处理硫代磷酰胺酯类衍生物杀菌作用的构效关系,并与逐步回归和偏最小二乘法作比较,模型的复相关系数R和剩余标准差Sy都较好.
The numerical genetic algorithm is combined with nonlinear PLS to improve and perfect the nonlinear PLS. The formulas of exponential, logarithm, reciprocal and Sigmoid functions are deduced. An algorithm that can deal with many nonlinear function relations is constructed, which can be used to solve complex Structure and performance interrelationships. The structure-activity relationship of the bactericidal effect of the thiophosphamide ester derivatives was treated by this method. Compared with the stepwise regression and partial least squares, the complex correlation coefficient R and the residual standard deviation Sy of the model were better.