Modelling and Forecasting of Monthly Crude Palm Oil Price of Malaysia using Hybrid Wavelet-Modified GMDH Model

Abstract

This study comprises developing a more appropriate hybrid wavelet-modified GMDH model for forecasting the monthly crude palm oil (CPO) price of Malaysia. In the proposed hybrid model, the complex data of monthly CPO price is decomposed into different sub series using discrete wavelet transform (DWT) and then it has been linked with modified GMDH model. Sigmoid, radial basis, tangent and polynomial functions are selected as transfer functions in modified GMDH for the best fit and correct model compared to conventional GMDH. The monthly CPO data were taken from Malaysian Palm Oil Board (MPOB) spanning the period January 1983 to November 2019. The capabilities of modified GMDH and hybrid wavelet-modified GMDH in modelling and forecasting the monthly CPO price are determined by MAE, RMSE, MAPE, R and R2. The MAPE of the proposed hybrid wavelet-modified GMDH model for the monthly CPO price of Malaysia is less than 4 % and coefficient of correlation (R) is 0.99, which show an excellent fit compared to the individual modified GMDH model. The proposed hybrid model provides the best alternative tools to help the Malaysian industry deal with price variations and assists Malaysia in playing a dominant role worldwide in the international market.

Keywords

• hybrid wavelet-modified gmdh model;
• discrete wavelet transform;
• cpo;
• modelling and forecasting;
• transfer functions