FORECASTING THE AMOUNT OF CURRENCY IN CIRCULATION AND THE VOLUME OF ELECTRONIC MONEY TRANSACTIONS USING A HYBRID VECTOR AUTOREGRESSIVE–GATED RECURRENT UNIT (VAR-GRU) METHOD

Zakiah Ulfa
Nusyirwan Nusyirwan
Dina Eka Nurvazly
Khoirin Nisa

RAGAM: Journal of Statistics and Its Application
Vol 5, No 2 (2026): RAGAM: Journal of Statistics & Its Application

Vector Autoregressive (VAR) is a method for forecasting multivariate time series data, but VAR is less able to represent nonlinear patterns in the data. One method in deep learning that is able to capture nonlinear patterns in data is the Gated Recurrent Unit (GRU). The hybrid VAR-GRU method combines both methods to complement each other. The data used in this study are the amount of currency in circulation and the volume of electronic money transactions from January 2019 to December 2024. The results show that the hybrid VAR-GRU has better performance than the single VAR model. The MAPE value obtained by the hybrid model for the volume of electronic money transactions is 4.39% and for the amount of currency in circulation is 2.93%, so the combined MAPE is 3.66%. Meanwhile, the VAR model produces a combined MAPE of 5.22%.