FORECASTING INDONESIA GOLD PRICE USING SETAR AND SETAR-TREE
Aurell Faza Ashilla(a*), Heri Kuswanto(a), Nur Iriawan(a)

a) Department of Statistics, Institut Teknologi Sepuluh Nopember (ITS), Sukolilo, Surabaya 60111, Indonesia
*aurellashilla[at]gmail.com

Abstract

Gold is used as a financial standard in many countries and also as a relatively enduring and accepted medium of exchange across countries. Gold was chosen as an investment instrument because it is liquid or easily converted into cash as a legal medium of exchange. In addition, the advantages of investing in gold are low risk, as a hedging tool, and prices that are not affected by interest rate policies. Even though the investment risk is low, not all gold investors can make a profit from the gold price. When viewed from the price movements that tend to fluctuate and have high volatility, the price of gold can contain a non-linear component. In order to accommodate nonlinear patterns in gold prices, nonlinear modeling is needed to predict gold prices in the future. The method used in this research is SETAR and SETAR-Tree to predict the price of Indonesian gold. Based on the results of the analysis carried out, in the in sample and out sample modeling, SETAR(2,1,1) and SETAR-Tree have almost the same performance because when compared through AIC values, the SETAR-Tree model has a smaller AIC value . Meanwhile, when compared through RMSE and MAPE values, the SETAR(2,1,1) model has smaller RMSE and MAPE values. In this study, a simulation was also carried out on the generated data following SETAR(2,2,2) with a different amount of data, namely 200 and 2000 data. Based on the simulation results, the SETAR and SETAR-Tree model has the same performance both for data with a small amount of data and a large amount of data.

Keywords: Gold Price, Nonlinear, Forecasting, Trees Regression, SETAR, SETAR-Tree

Topic: Mathematics and Statistics