Premium
An ARIMA‐ANN Hybrid Model for Time Series Forecasting
Author(s) -
Wang Li,
Zou Haofei,
Su Jia,
Li Ling,
Chaudhry Sohail
Publication year - 2013
Publication title -
systems research and behavioral science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.371
H-Index - 45
eISSN - 1099-1743
pISSN - 1092-7026
DOI - 10.1002/sres.2179
Subject(s) - autoregressive integrated moving average , artificial neural network , computer science , autoregressive model , time series , ibm , series (stratigraphy) , set (abstract data type) , nonlinear system , data set , artificial intelligence , machine learning , data mining , econometrics , mathematics , paleontology , materials science , physics , quantum mechanics , biology , programming language , nanotechnology
Autoregressive integrated moving average (ARIMA) model has been successfully applied as a popular linear model for economic time series forecasting. In addition, during the recent years, artificial neural networks (ANNs) have been used to capture the complex economic relationships with a variety of patterns as they serve as a powerful and flexible computational tool. However, most of these studies have been characterized by mixed results in terms of the effectiveness of the ANN s model compared with the ARIMA model. In this paper, we propose a hybrid model, which is distinctive in integrating the advantages of ARIMA and ANNs in modeling the linear and nonlinear behaviors in the data set. The hybrid model was tested on three sets of actual data, namely, the Wolf's sunspot data, the Canadian lynx data and the IBM stock price data. Our computational experience indicates the effectiveness of the new combinatorial model in obtaining more accurate forecasting as compared to existing models. Copyright © 2013 John Wiley & Sons, Ltd.