دراسة إحصائية مقارنة بين نماذج ARIMA و ANN : مفردة وهجينة (مع التطبيق)‎

Abstract EN

The research tackles constructing the hybrid models generated from integrated autoregressive moving average model (ARIMA) with the artificial neural networks (ANNs), This is done using ARIMA residuals as inputs for ANN model or vice versa, based on global gold price time series data to compare hybrid models of ARIMA-ANN and ANN-ARIMA with single models ARIMA and ANN, These models are compared using the RMSE and MAE prediction accuracy criteria to find the most appropriate model for predicting future values.

This study reveals the superiority of ARIMA-ANN over other models by having the lowest values of the prediction criteria's.