TY - GEN
T1 - Forecasting currency circulation data of Bank Indonesia by using hybrid ARIMAX-ANN model
AU - Prayoga, I. Gede Surya Adi
AU - Suhartono,
AU - Rahayu, Santi Puteri
N1 - Publisher Copyright:
© 2017 Author(s).
PY - 2017/5/12
Y1 - 2017/5/12
N2 - The purpose of this study is to forecast currency inflow and outflow data of Bank Indonesia. Currency circulation in Indonesia is highly influenced by the presence of Eid al-Fitr. One way to forecast the data with Eid al-Fitr effect is using autoregressive integrated moving average with exogenous input (ARIMAX) model. However, ARIMAX is a linear model, which cannot handle nonlinear correlation structures of the data. In the field of forecasting, inaccurate predictions can be considered caused by the existence of nonlinear components that are uncaptured by the model. In this paper, we propose a hybrid model of ARIMAX and artificial neural networks (ANN) that can handle both linear and nonlinear correlation. This method was applied for 46 series of currency inflow and 46 series of currency outflow. The results showed that based on out-of-sample root mean squared error (RMSE), the hybrid models are up to10.26 and 10.65 percent better than ARIMAX for inflow and outflow series, respectively. It means that ANN performs well in modeling nonlinear correlation of the data and can increase the accuracy of linear model.
AB - The purpose of this study is to forecast currency inflow and outflow data of Bank Indonesia. Currency circulation in Indonesia is highly influenced by the presence of Eid al-Fitr. One way to forecast the data with Eid al-Fitr effect is using autoregressive integrated moving average with exogenous input (ARIMAX) model. However, ARIMAX is a linear model, which cannot handle nonlinear correlation structures of the data. In the field of forecasting, inaccurate predictions can be considered caused by the existence of nonlinear components that are uncaptured by the model. In this paper, we propose a hybrid model of ARIMAX and artificial neural networks (ANN) that can handle both linear and nonlinear correlation. This method was applied for 46 series of currency inflow and 46 series of currency outflow. The results showed that based on out-of-sample root mean squared error (RMSE), the hybrid models are up to10.26 and 10.65 percent better than ARIMAX for inflow and outflow series, respectively. It means that ANN performs well in modeling nonlinear correlation of the data and can increase the accuracy of linear model.
UR - http://www.scopus.com/inward/record.url?scp=85019692180&partnerID=8YFLogxK
U2 - 10.1063/1.4982867
DO - 10.1063/1.4982867
M3 - Conference contribution
AN - SCOPUS:85019692180
T3 - AIP Conference Proceedings
BT - 3rd ISM International Statistical Conference 2016, ISM 2016
A2 - Abu Bakar, Shaiful Anuar
A2 - Mohamed, Ibrahim
A2 - Yunus, Rossita Mohamad
PB - American Institute of Physics Inc.
T2 - 3rd ISM International Statistical Conference 2016: Bringing Professionalism and Prestige in Statistics, ISM 2016
Y2 - 9 August 2016 through 11 August 2016
ER -