The Forecasting of Wholesale and Retail Prices of Rainbow Trout Fish Using Artificial Neural Network and ARMA Model

Paying attention to the relative stability of prices and forecasting prices could play an important role in controlling the instability of prices and ultimately reduce market risk. Comparison of different methods is important in the forecasting issues. In this study Wholesale and Retail Prices (Weekly Prices) of Rainbow Trout Fish will forecast with contrasting between forecasting power of the ARMA method and Artificial Neural Network method and choice the better one. In this study the Feed-forward network that is one of Back Propagation networks is used. The using data are from first week of farvardin 1388 to last week of shahrivar 1390. Before usage of predict methods، the random or non-randomized nature of the data were examined. Both of Price Series are predictable and non-randomized based on Random tests of Wald-Wolfowitz، Wallis-Moore and Durbin –Watson. Series are stationary in levels based on data stationary test (Dicky-Fuller augmented). Results of Forecasting show that in the model ARMA compared with artificial neural network (ANN) error rate is less based on four criteria forecast accuracy evaluating. Then it has higher power in forecasting the price of Rainbow Trout Fish. In the ANN model 80% of data for training network and 20% for testing network have been used. The results of accuracy equality test of the two methods (MGN) shows the ARMA model is also better than the neural network model in forecasting retail and wholesale prices significantly.