<![CDATA[This study proposes the application of the LSTM-Attention model to forecast the IDR exchange rate against the USD. Exchange rate stability is an important element in national and international economic resilience systems, as currency fluctuations can have a significant impact on trade, investment, banking, and household consumption. In the case of Indonesia, which is highly dependent on imported goods, exchange rate fluctuations cause an increase in import costs, rising inflation, and a decline in the competitiveness of export products in the global market, making accurate forecasting of exchange rate movements essential for economic policy, business strategy, and risk management. Statistical models such as ARIMA have been widely applied in exchange rate forecasting, but they have difficulty capturing the nonlinear of time series data. In recent years, machine learning methods such as Long Short-Term Memory (LSTM) have demonstrated their ability to handle timeseries data. Previous studies have shown that LSTM models generally outperform traditional methods, but they still face limitations in identifying important features across time steps. To overcome this problem, the Attention mechanism allows the model to focus on the most informative parts of the input sequence, thereby improving prediction accuracy. Experimental results show that the LSTM-Attention achieves MAPE of 1.28% and R2 of 0.97 and runtime 45% faster than BiLSTM. While BiLSTM achieved slightly higher accuracy, it’s required nearly twice the training time. Findings indicates that the proposed model offers practical choice for real-time exchange rate forecasting.]]>
