PERAMALAN NILAI TUKAR RUPIAH TERHADAP DOLLAR AMERIKA DENGAN METODE LONG SHORT-TERM MEMORY DAN GATED RECURRENT UNIT

Abstract

Indonesia is one of the countries predicted to experience an economic recession in 2023. Economic recession is a significant and sustained decline in economic activity, that can be observed through the weakening of the exchange rate. Information about future exchange rate changes can be obtained through forecasting. This research was conducted to obtain the best model for forecasting the exchange rate of Indonesian rupiah against US dollar using Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) methods. LSTM and GRU are built with memory cells, allowing them to store and access information better. This study uses data from 2 January 2018 – 14 February 2023. The selection of the best model was obtained through hyperparameter tuning. The research concluded that the LSTM model with 1 unit, 32 batch sizes, 50 time steps, and 314 epochs is the best model, as it achieved small values of MSE (0,001465) and MAPE (0,2321061%).