Short-Term Price Forecasting for Agro-products
Using Artificial Neural Networks

LI Ganqiong, XU Shiwei, LI Zhemin

Agricultural Information Institute of Chinese Academy of Agricultural Sciences

Abstract:
It is well known that short-term market price forecasting has been a difficult problem for a long time because of too many factors which cannot be accurately predicted in advance. Conventionally, time series analysis has been often employed in modeling short-term price forecasts. In recent ten years a relatively new technique of artificial neural networks has been proposed as an efficient tool for modeling and forecasting. The primary objective of this study attempted to examine the new technique of artificial neural networks for agro-products short-term price forecasting. The data used in this study consist of daily wholesale price, weekly wholesale price as well as monthly wholesale price since 1996. A feed-forward artificial neural network model and time series model (ARIMA) have been developed and compared. The conclusion is that the artificial neural network model evidently outperformed time series model developed in this study in forecasting ahead of one day or one week. An average absolute error under 5.0% in forecasting ahead of one day or one week was achieved from the feed-forward artificial neural network model, which also showed the good correlation between the modeled and observed price.

Key words: short-term forecasting, neural networks, regression analysis, time series analysis