ABSTRACTS OF ARTICLES OF THE JOURNAL "INFORMATION TECHNOLOGIES".
No. 4. Vol. 28. 2022

DOI: 10.17587/it.28.219-224

S. A. Tarasova, Ph.D, Senior Lecturer, Kursk State Medical University, Kursk, 305041, Russian Federation

Information Value Factor in Adaptive Time Series Forecasting

The article considers the task of forecasting time series with unstable dynamics. Adaptive forecasting methods are usually used to solve such tasks, but there is a problem of using adaptive forecasting methods, which consists of choosing an adequate adaptation parameter, in science. The purpose of the study is to construct and test an adaptive forecasting model, in which the adaptation parameter is calculated based on the information value. The discount factor is calculated as the ratio of the measure of the information value in the current period to the measure of the information value in the previous period. The values of the adaptation parameter and the discount factor are obtained depending on the information half-life. The analysis of the effectiveness of the model is carried out on the example of forecasting a medical and statistical indicator — the morbidity of the population. The average relative error of the forecast obtained in the study for the proposed adaptive model is significantly less than for the linear model. The method of finding the adaptation parameter and the discount factor based on the information value can serve as an additional criterion for choosing these constants, and in some cases the method of forecasting.
Keywords: time series, adaptive forecasting, adaptive forecasting methods, Brown's adaptive model, adaptation parameter, discount factor, information value, outdating of information, information half-life

P. 219–224

To the contents