Determining the Best-Fit Model for Oil Palm Yield and Planted Area in Malaysia

Abstract

Accurate information on agricultural production pattern is critical for foresighted and informed planning. Mathematical models have been successfully used to analyse and forecast a variety of agricultural data. In this study, time series models are employed for analysing and predicting the production and total planted area of oil palm, a significant agricultural crop export of Malaysia. Forecasting models, namely Linear Trend Model, Double Exponential Smoothing and Auto Regressive Integrated Moving Average (ARIMA) models are individually fitted to the 1974-2016 data to determine suitable models for forecasting purposes. Performances of the models are evaluated and compared using mean absolute percentage error (MAPE) and mean absolute deviation (MAD). The overall results demonstrated that ARIMA models are the best goodness-of-fit for both production and total planted area, indicating that there exists significant autocorrelation. In particular, ARIMA (2,2,3) is the best model to represent total planted area. Forecasted values of ten years show total planted land area will be gradually increasing, with an estimated increase of 0.14% to 1.98% per annum. Meanwhile, ARIMA (0,1,1) has the lowest MAPE and MAD value, suggesting that the model is most appropriate to represent oil palm production. The forecasted values show a consistent increase of about 0.12% per annum for the next ten years. The predicted results in this study could be used by relevant parties such as policy makers to foresee ahead of time the future requirement of import/export of palm production in Malaysia and enable them to do better strategic planning. This would indirectly contribute towards strengthening the position of oil palm industry in Malaysia and the world.