ARIMA MODEL IN MACHINE LEARNING: TIME SERIES PREDICTION

UDC 004.62
DOI: 10.36871/2618-9976.2024.01.004

Authors

Mariya V. Dobrina,
Financial University under the Government of the Russian Federation, Moscow, Russia

Abstract

The paper considers the ARIMA method as a time series forecasting tool. Most often, its use is aimed at checking their stationarity. Usually, special tests of unit roots and the order of integration of the series are used. Then (when the order of integration exceeds zero), the series is transformed by searching for the difference in the required order. In the completed work, the process is illustrated by building on the example of the ticker of the OZON company. A graph of time series components, autocorrelation and partial autocorrelation functions were formed. The study resulted in forecasts for OZON, Alrosa and VTB, built with the help of ARIMA, as well as a comparison of these forecasts.

Keywords

ARIMA approach, Linear forecast, Time series, Correlogram