Page 300 - Contributed Paper Session (CPS) - Volume 7

CPS2105 Hermansah et al.

Comparison of ARIMA, Neural Network and
Wavelet Models for Forecasting Indonesia Sharia
Stock Index
3
3
1,2
Hermansah ; Dedi Rosadi ; Herni Utami ; Abdurakhman
3
1 Ph. D. Student of Mathematics, Universitas Gadjah Mada, Yogyakarta, Indonesia
2 Department of Mathematics Education, Universitas Riau Kepulauan, Batam, Indonesia
3 Department of Mathematics, Universitas Gadjah Mada, Yogyakarta, Indonesia

Abstract
Forecasting is an activity that predicts what happens in the future based on
the present and past values of a variable. Forecasting is a very important
element, especially in planning and decision making. The method that is often
used in data forecasting that may occur in the future is the Autoregressive
Integrated Moving Average (ARIMA). In the case study, a comparison of
forecasting models is made using the ARIMA, Neural Network and Wavelet
methods. The Neural Network (NN) method used is Feed-Forward Neural
Network (FFNN) or often called Back-propagation Neural Network (BPNN) and
Wavelet used is the Daubechies 4 with wavelet type Maximal Overlap Discrete
Wavelet Transform (MODWT). Based on the case study on the data close of
Indonesia Sharia Stock Index (ISSI), the MSE value obtained of forecasting the
ARIMA method is 4,846185 and MAPE is 0,011158. MSE of forecasting the NN
method is 2,419994 and MAPE is 0,007553. MSE of forecasting the Wavelet
method is 38,620430 and MAPE is 0,032779. Therefore, for forecasting ISSI
close data it can be said that the best model is the NN model because the MSE
and MAPE values obtained are smaller than the ARIMA and Wavelet models.

Keywords
Forecasting Indonesia Sharia Stock Index; ARIMA; Neural Network; Wavelet

1. Introduction
Forecasting is an activity that predicts what happens in the future based
on the present and past values of a variable [1]. Forecasting is a very important
element, especially in planning and decision making. The grace period
between an event and the upcoming event is the main reason for forecasting
and planning. In these situations forecasting is an important tool in effective
and efficient planning. The choice of method in forecasting depends on
several aspects of research, namely aspects of time, data patterns, types of
system models observed, and the level of accuracy of forecasting. The use of
these methods in forecasting must fulfill the assumptions used [2].
This study uses three forecasting methods, namely Autoregressive
Integrated Moving Average (ARIMA), Neural Network (NN) and Wavelet.
ARIMA often called the Box-Jenkins time series method is a method that uses

287 | I S I W S C 2 0 1 9