Gold demand forecasting in Malaysia: A Hybrid Nar-Arima Approach

Abstract

This study investigates the demand for gold in Malaysia from 2010 to 2023, utilizing both nonlinear autoregressive (NAR) and autoregressive integrated moving averages (ARIMA) models to assess their predictive performance. Numerous global and domestic factors, such as market conditions, investor behavior, and economic fluctuations, influence gold demand, a critical economic indicator. We applied the NAR model, known for its ability to capture complex nonlinear relationships, and the ARIMA model, recognized for modelling linear trends, individually and in combination to forecast future gold demand. The results show that ARIMA performs well in capturing linear trends in the data, while NAR effectively captures nonlinear patterns. We proposed a hybrid NAR-ARIMA model that combines the strengths of both approaches, leading to enhanced forecast accuracy. The analysis highlights the effectiveness of this hybrid model in providing more reliable predictions, particularly for datasets exhibiting both linear and nonlinear behaviors. We then used the hybrid model to forecast gold demand in Malaysia up to 2030, providing valuable insights for policymakers, investors, and stakeholders in the gold market. These findings contribute to the growing body of literature on advanced time series forecasting techniques and their applicability to commodity demand forecasting.