Using Neural Networks for USD to IDR Exchange Rate Prediction

Abstract

Predicting the USD to IDR exchange rate is critical for financial markets, international trade, and economic policy. This research employs neural networks to model the complex and non-linear patterns inherent in time-series data. The methodology involves collecting historical daily exchange rate data, preprocessing to handle missing values, normalizing features, and transforming the data into a format suitable for modeling. The neural network architectures utilized include Feedforward Neural Networks (FNN), Long Short-Term Memory (LSTM), and Gated Recurrent Units (GRU). Model evaluation metrics, such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE), indicate the neural networks’ effectiveness in capturing general trends with high accuracy, despite challenges during periods of high market volatility. Comparative analysis with traditional methods, such as ARIMA, highlights the superior ability of neural networks to manage non-linear relationships and long-term dependencies. This study provides valuable insights into developing advanced tools for exchange rate prediction, leveraging the power of machine learning. The results demonstrate the potential of neural networks in financial forecasting, with opportunities for improvement through integrating additional external factors and optimizing model architectures.