Russian Federation
Computer models of time series forecasting are considered. The forecasting models are implemented in the form of computer programs. The results of computational experiments on estimating the error of short-term forecasting of time series are presented. Mathematical models used to solve forecasting problems are considered, including self-organizing models, models in the form of fuzzy inference systems, models of multilayer feed-forward neural networks, adaptive forecasting models, and piecewise polynomial approximation models. The main attention is paid to short-term forecasting of time series, in which forecasting is carried out one time interval ahead. As computer models of short-term forecasting, a model of an artificial multilayer neural network without feedback with a linear activation function, an exponential smoothing model with adaptation at each time step are considered in detail to time series data and a model of piecewise polynomial approximation, in which the approximating function is composed of individual polynomials of the same small degree (of the third degree – cubic splines). Each computer model is implemented as a computer program, for which a block diagram and program interface are given.
computational experiment, short-term forecasting, time series, mathematical model, computer simulation, computer program
1. Mathematical modeling of complex technical systems: sat. articles. M.: MSTU im. Bauman, 1997.
2. Labinsky A.Yu., Podruzhkina T.A. Reduction of technogenic risks with predictive mathematical models // Natural and technogenic risks. 2013. № 3.
3. Kahneman D. Mathematical Modeling. New York and London, 2016.
4. Pegat A. Fuzzy modeling and control. M.: BINOM, 2013.
5. Labinsky A.Yu. Model of fuzzy forecasting // Problems of risk management in the technosphere. 2016. № 4.
6. Mamdani E.H., Assilian S. An experiment with a fuzzy logic controller // Cybernetics and Systems. 2014. № 15.
7. Tarkhov D.A. Neural networks as a means of mathematical modeling. M.: Radio engineering, 2006.
8. Labinsky A.Yu. Peculiarities of using a neural network for forecasting time series // Scientific and analytical journal «Vestnik Saint-Petersburg university of State fire service of EMERCOM of Russia». 2018. № 1.
9. Barron A.R. Neural net approximation // IEEE Transaction on Information Theory. 2016. Vol. 49.
10. Baldi P. A recurrent neural network // Neural Computation. 2018. Vol. 2.
