Performance Comparison of LSTM and NHITS for Forecasting the Indonesian Rupiah–US Dollar Exchange Rate

The exchange rate of the rupiah against the US dollar plays a crucial role in Indonesia’s economy. Exchange rates fluctuate nonlinearly according to market prices, creating challenges for accurate forecasting. Various machine learning methods address these challenges, including Long Short-Term Memory (LSTM) and Neural Hierarchical Interpolation Time Series (NHITS). Therefore, this study compared the performance of LSTM and NHITS models in forecasting the exchange rate of the rupiah against the US dollar. Daily data from January 2, 2017, to January 30, 2026, from Yahoo Finance were used. The analysis included data exploration, preprocessing, LSTM modelling, NHITS modelling, performance comparison, and forecasting. The results indicated that NHITS produced slightly better forecasting results than LSTM under the evaluation setting used in this study, with Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Pearson correlation values of 1.12%, 87.39, 233.11, and 0.85, respectively, along with shorter computation time. Meanwhile, LSTM produced MAPE, MAE, RMSE, and Pearson correlation values of 1.24%, 211.38, 238.93, and 0.80, respectively. The forecasts showed that the NHITS model better captured actual data changes, while the LSTM model produced smoother patterns with smaller fluctuations than those observed in actual data.