Nonparametric regression models for time series analysis and forecasting 

Aleksey V. Kostin, Head of Zarechny of Penza region (27 30-letiya Pobedy avenue, Zarechny, Penza region, Russia) E-mail: ekostin@obl.penza.net
Petr P. Makarychev, Doctor of engineering sciences, professor, professor of the sub-department of mathematical support and application of computers, Penza State Univesity (40 Krasnaya street, Penza, Russia), E-mail: makpp@yandex.ru

Background. Regression analysis is a type of machine learning. With the application of regression analysis, the problems of structural-parametric identification, and predicting the behavior of systems and objects are solved. Regression models constructed using observed data at finite time intervals are time series models. The purpose of the study, the results of which are presented in the article, is to develop non-parametric regression models for the analysis and forecasting of fires, tragic events on the water, accidents on water pipes, road traffic accidents in the region. Materials and methods. Analysis and forecasting of time series levels reflecting emergencies and events in the region, by machine learning using nonparametric regression models based on linear and nonlinear functions of activation of artificial neurons. Results.The analysis and prediction of the time series levels are set and solved. Content of the problem: analysis of the stagnation of the
time series; elimination of random emissions; allocation of the piece-line trend; development of non-parametric models of machine learning; performance of a omputational experiment and evaluation of the quality of forecasting. Conclusions. The results of the computational experiments confirmed the prospects of applying the machine learning method using nonparametric regression models based on Fourier functions, piecelinear and nonlinear functions.

the time series, trend, machine learning model, non-parametric regression model, forecasting, an assessment of the quality of forecasting

Kostin A.V., Makarychev P.P. Nonparametric regression models for time series analysis and forecasting. Izvestiya vysshikh uchebnykh zavedeniy. Povolzhskiy region. Tekhnicheskie nauki = University proceedings. Volga region. Engineering sciences. 2024;(4):16–27. (In Russ.). doi: 10.21685/2072-3059-2024-4-2