توظيف الشبكات العصبية الاصطناعية في التنبؤ لنماذج الانحدار الذاتي و المتوسطات المتحركة بمتغيرات خارجية المنشأ

Abstract EN

This study aims to compare an efficiency between two methods in prediction , time series and artificial neural networks (ANN).By using one of measures of efficiency of the model which is mean square errors (MSE) also,the coefficient of determination (R), it was noted , in this study, that the model of artificial neural networkswasmore efficiency than Box-Jenkinsmodel in prediction .

Themodel of Autoregressive integrated for the moving average with exogenous variables ARIMAX(p,d,q) is the best dynamic models which are used in a prediction to a value for any known phenomenon because the mathematic shape for thismodel which is built by mixing two models; Autoregressive and moving average , also thismodel is the best because it gives a good efficiency results in prediction more than if each model is built alone .

In addition , thismodel has an attribute that is not linear in parameters , thus it is estimated by non-linear minimum square (NLS) .

The study aimed to the comparative between the results of artificial neural networks model and the results of Autoregressive integrated for the moving average with exogenous variables model in prediction about the relationship between the temperature, Pressure as inputs and CO as output was resolved in the soft drinks (7- up), the results reached that the variance of prediction errors for the ANN model was less than variance of prediction errors for ARIMAX model .