FORECASTING TIME SERIES WITH COMPLEX BEHAVIOR USING THE GENERALIZED TRIGONOMETRICAL MODEL WITH RANDOM PARAMETERS

The possibility of using different methods of modeling the time series of quarterly GDP values of Ukraine, in particular, autocorrelation models with different sets of parameters, was investigated. For this reason, 16 models were built and their quality was tested. The SARIMA(4,1,1)∙(0,0,1)_4 model was studied and used to forecast the values of the time series, and the predictive accuracy was estimated. A generalized trigonometric model with random components has been developed for modeling a series of first differences taking into account random perturbations. The obtained model was applied to the analysis of Ukraine's GDP indicators, forecasting was performed according to two scenarios: the pessimistic and the most expected, and the forecasting results were compared with empirical data. It is shown that this model can be effectively used for modeling and forecasting some time series with random disturbances. Keywords: forecasting time series, time series stationarity, ARIMA and SARIMA models, trigonometric models, simulation modeling.