The expert system of search the forecasting method with using of neural network in volatility conditions of initial data

The problem of a choice of a method of forecasting in view of statistical characteristics of the initial data is considered in this work. The concept of construction of expert system of a choice of a method of forecasting of time numbers in volatility conditions the initial data is developed. Statistical characteristics on the basis of which the choice of a method of forecasting in heteroskedacity conditions is possible, and also criterion of efficiency for comparison of methods of forecasting are defined. To be under construction imitating model. The neural network for search of an optimum method of forecasting which has been trained by means of the data received on the basis of imitating modeling is offered and under construction.