微分方程演化建模用于色谱保留时间的研究

Evolutionary Modeling of Differential Equations for the Study of Chromatographic Retention

利用演化算法的自适应、自组织、自学习的特性,设计了遗传程序设计与遗传算法相嵌套的混合演化建模算法,以遗传程序设计优化模型结构,以遗传算法优化模型参数,为化合物的液相色谱容量因子随流动相组成变化关系自动建立微分方程演化模型。通过对7个化合物的建模结果表明,演化模型的拟合和预测精度均明显高于常规的GM(1,1)模型和改进的GM(1,1)模型。

Based on the properties of self-adaptation, self-orgainization and self-learning of evolutionary algorithm, this paper proposes a hybrid evolutionary modeling algorithm to build up models of differential equations for the dependence of capacity factors of compounds on mobile phase automatically. The main idea was to embed genetic algorithm (GA) in genetic programming (GP) where GP was employed to optimize the structure of a model while GA was employed to optimize its parameters. The experimental results of seven compounds showed that the fitting and prediction accuracy of evolutionary models obtained by using the algorithm was much higher than that of GM( 1,1) models and modified GM( 1,1) models.