Forecasting IDR Exchange Rate to USD Using Hybrid ARIMA – LSTM

Zulkifli Rais (Universitas Negeri Makassar)
Sitti Masyitah Meliyani R (Universitas Negeri Makassar)
Astrid Suwardani Sumarno (Universitas Negeri Makassar)
Agung Tri Utomo (ARRUS Journal)

ARRUS Journal of Engineering and Technology
Vol. 6 No. 1 (2026)

Time series forecasting often involves both linear and nonlinear patterns, making the use of a single method less effective. This study aims to forecast the exchange rate of the Indonesian Rupiah (IDR) against the United States Dollar (USD) using a hybrid ARIMA–LSTM model. ARIMA is used to capture linear patterns, while LSTM is employed to model nonlinear residual components. The data used are weekly exchange rates from January 2020 to August 2025. Model performance is evaluated using Mean Absolute Percentage Error (MAPE). The results show that the hybrid ARIMA–LSTM model produces better forecasting accuracy compared to individual ARIMA and LSTM models, with the lowest MAPE value of 0.73%. This indicates that combining linear and nonlinear modeling approaches improves forecasting performance for complex time series data.