A Kernel Grey Model with Genetic Algorithm Optimizer and its applications in forecasting the palm oil price in Malaysia

Accurate forecasting is difficult since palm oil prices are consistently highly nonlinear. It is important to choose the right forecasting models since there are several available. The grey model has proven to be a good forecasting model. Nevertheless, the majority of extant grey models are fundamentally linear models, which limits their ability to capture nonlinear trends. This paper introduces a nonlinear extended parametric grey model known as the kernel grey model (KGM). However, the prediction of the KGM model is dependent on the kernel function and the KGM parameters. This study presents a genetic algorithm-based enhanced KGM model and verified it with data on Malaysia's palm oil prices from 2000 to 2019. The multivariable linear regression (MLR) and support vector machine (SVM) forecasting models were chosen for comparison based on mean absolute percent error and root mean square percent error. The results reveal that KGM outperforms the other two models in training and testing data performance, and it can significantly enhance forecast accuracy.