Empirical Mode Decomposition Based Ensemble Hybrid Machine Learning Models for Agricultural Commodity Price Forecasting

Abstract

Agricultural commodity price is very volatile in nature due to its nonlinearity and nonstationary character. The volatility behaviour of the commodity price creates a lot of problems for both producer and consumer. The steady forecast of the price may reduce the problems and increase the profit for the stakeholders. In this study, an ensemble hybrid machine learning model based on empirical mode decomposition (EMD) has been proposed to forecast the commodity price. EMD decomposes the nonstationary and nonlinear price series into different stationary intrinsic mode functions (IMF) and a final residue. Then Machine learning techniques like Artificial neural network (ANN) and Support vector regression (SVR) were used to forecast each of the decomposed components. Finally, all the forecasted values of the decomposed components were aggregated to produce the final forecast. Two R modules were developed for the application of the proposed methodology. The proposed methodology has been applied to the monthly wholesale price index of vegetables. The results indicated that the ensemble hybrid machine learning model based on empirical mode decomposition has superior performance compared to generic models