SOFTWARE FOR SEQUENTIAL IDENTIFICATION OF NON-STATIONARY TIME SERIES

This paper describes the features of the software implementation of an algorithm for time series model identification. The algorithm that based on different probabilistic and statistical methods (singular spectrum analysis, ARIMA, correlation and regression analysis) has been implemented as a plugin for the client-server version of the forecasting system VarForecasts. This plugin was tested using time series of water inflows to the Novosibirsk Hydroelectric Power Station site. A comprehensive model of the described inflow time series was built by using the developed plugin. The model is adequate to the empirical data and can be used for daily inflow forecasting.