Forecasting Potato Price using Ensemble Artificial Neural Networks

Agricultural price information needs for decision-making at all levels are increasing due to globalization and market integration. Agricultural price forecasting is one of the challenging areas of time series analysis due to its strong dependence on biological processes. In this paper, an empirical mode decomposition based neural network model is employed for potato price forecasting. The daily potato wholesale price series from Delhi market was decomposed into eight independent intrinsic modes (IMFs) and one residue with different frequencies. Then anartificial neural network with single hidden layer was constructed to forecast these IMFs and residue component individually. Finally, the prediction results of all IMFs including residue are aggregated to formulate an ensemble output for the original price series. Empirical results demonstrated that the proposed ensemble model outperforms a single model in terms of root mean square error and directional prediction statistics.