Comparison of classical method and optimization methods for estimating parameters in nonlinear ordinary differential equation

Abstract

This study is concerned with the estimation of constant and time-varying parameters in non-linear ordinary differential equations, which do not have analytical solutions. The estimation is done in a multi-stage method where constant and time-varying parameters are estimated in a straight sequential way from several stages. In the first stage, the model of the differential equations is converted to a regression model that includes the state variables with their derivatives and then the estimation of the state variables and their derivatives in a penalized splines method and compensating the estimations in the regression model. In the second stage, the pseudo- least squares method was used to estimate the constant parameters. In the third stage, the remaining constant parameters and the time-varying parameters are estimated by using a semi-parametric regression model. This method is compared with the optimization method, which depends on the algorithm of differential evolution algorithm to estimate unknown parameters. The comparison was made using simulations. The results showed that the results were better to the method based on the differential evolution algorithm.