The ARIMA versus Artificial Neural Network Modeling
Khorshid, M., Tharwat, A., Bader, A., Omran, A. (2009). The ARIMA versus Artificial Neural Network Modeling. EKB Journal Management System, 2(1), 30-40. doi: 10.21608/ijci.2009.33936
Motaz Khorshid; Assem Tharwat; Amer Bader; Ahmed Omran. "The ARIMA versus Artificial Neural Network Modeling". EKB Journal Management System, 2, 1, 2009, 30-40. doi: 10.21608/ijci.2009.33936
Khorshid, M., Tharwat, A., Bader, A., Omran, A. (2009). 'The ARIMA versus Artificial Neural Network Modeling', EKB Journal Management System, 2(1), pp. 30-40. doi: 10.21608/ijci.2009.33936
Khorshid, M., Tharwat, A., Bader, A., Omran, A. The ARIMA versus Artificial Neural Network Modeling. EKB Journal Management System, 2009; 2(1): 30-40. doi: 10.21608/ijci.2009.33936

The ARIMA versus Artificial Neural Network Modeling

IJCI. International Journal of Computers and Information
Article 3, Volume 2, Issue 1, June 2009, Page 30-40  XML PDF (307.44 K)
Document Type: Original Article
DOI: 10.21608/ijci.2009.33936
View on SCiNiTO View on SCiNiTO
Authors
Motaz Khorshid* 1; Assem Tharwat1; Amer Bader1; Ahmed Omran2
1Faculty of Computer & Information. Cairo University, Egypt
2Central Lab For Agricultural Expert Systems. Agriculture Research Center, Egypt
Abstract
Linear models almost reach their limitations with non-linearity in the data. This paper provides a new empirical evidence on the relative macroeconomic forecasting performance of linear and nonlinear models. The well established 
and widely used univariate Auto-Regressive Integrated Moving Average (ARIMA) models are used as linear forecasting models whereas Artificial Neural Networks (ANN) are used as nonlinear forecasting models. The neural network paradigm that was selected for developing the proposed model is a Multi-layer Feedforward network based upon the Backpropagation training algorithm. ANN has been proven to be successful in handling nonlinear problem optimization and prediction. The forecasting models used to identify whether action is needed to alter the future, when such action should be taken by the decision maker in order to change the future of the bank or its environment to improve the bank's chance of achieving its targets. We applied the proposed model on a Financial Balance Sheet’s data of a commercial bank in Egypt. The Results show that, the proposed model (which dependent on the ANN) is more accurate than the other models, which depend on the ARIMA model with accuracy between 8 % and 10.4 %.
Keywords
Economic Forecast; Neural Networks; ARIMA; Back-propagation; and Time Series Models
Statistics
Article View: 242
PDF Download: 566