Forecasting comparisons using a hybrid ARFIMA and LRNN models

ABSTRACT

In this article, an autoregressive fractionally integrated moving average model (ARFIMA) and a layer recurrent neural network (LRNN) were combined to form a hybrid forecasting model. The hybrid model was applied on the daily crude oil production data of the Nigerian National Petroleum Corporation (NNPC) to forecast the daily crude oil production of the NNPC. The Bayesian model averaging technique was used to obtain a combined forecast from the two separate methods. A comparison was made between the hybrid model with standalone ARFIMA and LRNN methods in which the hybrid model produced better forecasting performance than the standalone methods.

KEYWORDS:

• Bayesian model averaging
• Autoregressive
• Neural network
• Mean absolute error
• Root mean square error and forecasting