Identifying and Forecasting the Factors that Derive CPO Prices in Malaysia Using NARX Model

Abstract

Determining the movement of the crude palm oil prices (CPO) is a crucial issue, whereby it always associated with the decisionmaking by businessman, investors, speculators and policymaker. Besides the CPO prices indicate that it is fluctuating all over a time and need to be forecast as to make it visible for the businessman, investors and policymaker in decision-making. The issues arises when most of the literature (1) relying too much on the historical data, whereby its predict series of CPO prices yt+1 given d past values of y(t), (2) disregarded the important factors that also derives the CPO prices, and (3) depending too much on univariate time series forecasting. Therefore, this study considered soybean price, export of palm oil, rainfall, and palm oil stock level as the factors or inputs that derive the CPO prices in Malaysia. We applied the Autoregressive Distributed Lag (ARDL) as an effort to check the long run relationship with the listed factors. In predicting the CPO prices, we employ the Nonlinear Autoregressive with External (NARX) with three different training algorithms that are Levenberg-Marquardt, Bayesian Regulation and Scaled Conjugate Gradient. The general findings demonstrated that three of the algorithms using the listed inputs show decent results for CPO prices prediction. Therefore, the listed inputs should not be disregarded as this study confirmed that its influence the CPO prices in Malaysia.