A comparative study on univariate time series based crude palm oil price prediction model using machine learning algorithms

Abstract

Crude palm oil (CPO) price prediction plays an important role in the agricultural economic development. It requires an in-depth knowledge in both economics and agricultural domain. The aim of this paper is to propose a CPO price prediction model to help the plantation organizations in the palm oil sector to effectively anticipate CPO price fluctuations and managing the resources more effectively. The CPO price behavior are non-linear in nature, thus prediction is very difficult. In this paper, a recurrent network, Long Short Term Memory (LSTM) based CPO price prediction system is compared with artificial neural network (ANN) and Holt-Winter method. The findings of this study shows that the LSTM based forecasting model outperformed other models in forecasting the CPO price movement. This study recommends that a LSTM based forecasting could better help the farmer and planters in the agriculture sector in managing the demand of CPO and the operation processes for a better return on investment.