مقايسه ي مدل هاي شبكه هاي عصبي مصنوعي و سري هاي زماني براي پيش بيني قيمت گوشت مرغ در ايران

A Comparison between Artificial Neural Networks and Time Series Models in Chicken Price Forecasting in Iran

Due to the importance of price prediction in providing protein products including chicken, this research uses methods of time series models (ARIMA) and artificial neural networks (ANNs) to forecast the trends of the chicken price in Iran for the various time paths consisting one, six and twelve months, respectively. Accordingly, we investigate the hypothesis in which ANNs are more efficient in chicken price prediction than time series models in doing so. Monthly data are collected for the period 1992:3-2007:2 (1371:1-1385:11), obtaining from the domestic resources related to the agricultural sector. The empirical results confirm that the performance of Back-Propagation Network, which is a type of ANNs, in forecasting the price trend has been mostly more precise than that of ARIMA. Also another ANN-based method, the Elman Network, has more efficiency in one month and twelve months time paths prediction than that of ARIMA. The implication is thus the use of ANNs in prediction of the Iranʹs chicken price could assist policymakers in the poultry industry to set up appropriate price policies through predicting correctly market fluctuations.