Mathematical simulation of dynamics on the basis of analysis of multidimensional time series with consideration for lagged influence of factors using neural networks

Abstract

This article investigates into the models and methods of neural simulation of dynamics on the basis of analysis of multidimensional time series with consideration for lagged influence of significant factors. Mathematical formulation of the neural network construction for nonzero lag is presented, peculiarities of lag optimization are described for one independent variable (input), simulation and forecast database is specified as well as neural algorithms of data processing, algorithmization of multivariate regression analysis is carried out with optimization of lag vector for significant factors.