A DUAL METHODS APPROACH TO CRUDE PALM OIL PRICE FORECASTING IN MALAYSIA: INSIGHTS FROM ARDL AND LSTM

Abstract

Understanding the volatile nature of palm oil prices is crucial due to its significant implications for the economy and the market. Due to its complexity, the central issue of the rise in palm oil price determinants and forecasting depends on various market demand and supply forces. However, many scholars fail to conclude that the factor drives palm oil prices. This study examines the factors affecting Malaysian Crude Palm Oil (CPO) pricing dynamics and uses estimated palm oil prices in forecasting. Using data from the Malaysian Palm Oil Board, spanning January 2004 to December 2021. Methodologically, we employed Autoregressive Distributed Lag (ARDL) and Long Short-Term Memory (LSTM) models to evaluate and forecast CPO prices. Our findings revealed that the LSTM model outperformed the ARDL model in forecasting accuracy. Notably, the LSTM model was more effective with a selection of ten independent variables identified through LASSO and SHAP estimation, compared to using either eleven or four variables based on ARDL regression results. The analysis highlights the significant influence of weather conditions and macroeconomic factors, particularly tax rates, on CPO prices. The findings enhance understanding of market dynamics and assist in accurate forecasting of CPO prices.