Authors: Augustine Pwasong; Saratha Sathasivam
Addresses: School of Mathematical Sciences, Universiti Sains Malaysia, 11800 Pulau Pinang, Malaysia ' School of Mathematical Sciences, Universiti Sains Malaysia, 11800 Pulau Pinang, Malaysia
Abstract: In this paper, a cascade forward back propagation neural (CFBN) network model and an ordinary least squares (OLS) regression model are fused together to form a hybrid model called 'hybrid OLS-CFBN' technique. The hybrid model was applied on the crude oil data of the Nigerian national petroleum corporation (NNPC) to determine the forecasting performance of the model. The fusion was made by the Bayesian model averaging (BMA) technique, to obtain a combined forecast from the two separate methods, that is, the CFBN and the OLS methods. The fusion produced the hybrid OLS-CFBN model which was used on the difference and log difference series of the NNPC data. The results indicate that the combined forecast have better forecasting performance greater than the standalone methods on the difference series based on the mean square error sense. The analysis for this study was simulated using MATLAB software, version 8.03.
Keywords: BMA; Bayesian model averaging; artificial neural networks; ANNs; MAE; mean absolute error; root mean square error; RMSE; time series forecasting; OLS regression; ordinary least squares; nonlinear; crude oil production; Nigeria; simulation; Nigerian National Petroleum Corporation; NNPC.
International Journal of Intelligent Systems Technologies and Applications, 2016 Vol.15 No.3, pp.255 - 280
Received: 06 Apr 2015
Accepted: 09 Feb 2016
Published online: 02 Aug 2016 *