Information and Intelligent Forecasting Systems Based on the Methods of Neural Network Theory

This paper is dedicated to the problem of the research of the information and intelligent forecasting systems which solve the problems of regression analysis. This class of the systems includes the systems that use different mathematical apparatus for forecasting, for the implementation of which a wide range of methods and algorithms is used. The paper also deals with the problems of regression analysis using elements of neural network theory. Developing a set of functions to form an approximating function allows you to create models that work efficiently. In existing information and intelligence systems, either implemented functions adapted to solve specific forecasting tasks, or, as in the case of neuropackages, it is considered that the proposed functions can be used to solve a wide range of tasks, adaptation to the conditions of the task occurs during the design of the neural network architecture. To create an approximating function, it is possible to implement linear and non-linear functions, which have proven themselves well in neural network analysis for solving a wide range of tasks and are obtained within the framework of deterministic chaos theory. To implement the function that approximates the dependence of the predicted value on the forecast arguments, the use of a two-layer heterogeneous, artificial neural network without feedback is proposed.