Mitigating Risks: A Hybrid Autoregressive Integrated Moving Average-Artificial Neural Network (ARIMA-ANN) Methodology for Exchange Rate Volatility

This study aims to estimate the rupiah exchange rate against the US dollar by employing a hybrid ARIMA-Artificial Neural Network (ARIMA-ANN) methodology, with export treated as an exogenous variable. It evaluates the precision of the model against a non-hybrid model. Multiple types of research have demonstrated the efficacy of the hybrid ARIMA-ANN model in minimizing errors, thereby justifying its selection. The hybrid ARIMA-ANN methodology employs ANN to discern nonlinear patterns in time series data and ARIMA to detect linear patterns. The results of this research indicate that the hybrid ARIMA-ANN model yields more precise forecasts. The RMSE value of 0.025 contrasts with the RMSE of 0.045 for the ARIMA model and 0.035 for the ANN. The significance of projecting exchange rate volatility holds both practical and scholarly value. Our study offers new insight by thoroughly analyzing the predictive capacity of financial and macroeconomic variables related to future exchange rate volatility.