A Comparison Of Feed Forward Neural Network Models And Time Series Models For Forecasting Turkey's Monthly Dairy Exports To Iraq

Abstract

Forecasting is a major branch of statistics with several applications, particularly in econometrics. Many governments utilize it to develop long-term goals and make future decisions. The two main forecasting approaches are examined in this paper to discover the best forecasting model for the monthly amount of dairy products exported from Turkey to Iraq. The Autoregressive Integrated Moving Average (ARIMA) model is used in the first technique, known as Box-Jenkins, while the Feed Forward Neural Network (FFNN) model is used in the second. The data, which comes from the official websites of the UN Comtrade and the Turkish Statistical Institute (TUIK), contains the monthly volume of dairy products exported between 2010 and 2019. For analysis, three software tools Alyuda NeuroIntelligence, R, and SPSS were used. This comparison also included Akaike Information Criteria (AIC), Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R2. According to the results, the FFNN model fits better than the ARIMA model. Furthermore, the FFNN model exhibits less errors than the ARIMA model and is much better in terms of goodness of fit due to lower MAE, RMSE, and AIC values.