Hybrid Quantile Regression Neural Network Model for Forecasting Currency Inflow and Outflow in Indonesia

Abstract

Regression analysis which can explain the relationship between variables on various quantiles has been developed using quantile regression. Moreover, quantile regression can be applied in forecasting analysis. The aim of this study was to find the best model for forecasting inflow and outflow in Indonesia which contains heteroscedasticity and nonlinearity problems. In order to improve the forecast accuracy, quantile regression will be combined with neural network method, known as quantile regression neural network (QRNN). Then, the forecast accuracy of QRNN will be compared with ARIMAX method based on some forecast accuracy criteria, i.e. RMSE, MAE, MdAE, MAPE, and MdAPE. Two types of data are used as case studies in this research, i.e. simulation and real data about 6 currencies of inflow and outflow in Indonesia. The result of simulation study shows that QRNN is the best method to solve heteroscedasticity and nonlinearity problem. Furthermore, the comparison results on real data shows that QRNN yield better result than ARIMAX for four currencies.