Analysis of forecasting estimates quality using the method of complexation

Abstract

Using estimates of short-term forecasts, obtained by different methods, complexation of forecasts estimates is done for one step. While on calculated RMS errors for each method (exponential smoothing, crisp and fuzzy MGAA, Kalman filter) scheme of association of forecasts estimates, which increases the quality of forecasting in conditions of correct choice of the weight factors is realized. The obtained experimental results using the above mentioned methods suggest opportunities to improve the quality of forecasts estimates through the use of the proposed scheme of combining in conditions when the dispersion errors of the estimates of individual forecasts are close to each other. To compare the results of forecasting a set of statistical parameters of the forecast quality is used.