Modelling of Forecasting ASEAN-5 Stock Price Index Using GSTAR Model

Abstract

This research aims to apply the Generalized Space-Time Autoregressive (GSTAR) model to predict stock price indices in ASEAN-5 countries. Generalized Space Time Autoregressive (GSTAR) model is one of the most common used space-time model to modeling and predicting spatial and time series data. The GSTAR model produces a space-time model that adopts the stages of the Autoregressive Integrated Moving Average (ARIMA) model. This research uses parameter estimation using the Maximum Likelihood method, which is a method used to estimate parameter values by maximizing the probability function seen based on observations. This research uses secondary data in the form of Stock Price Index data from 5 countries in Asia, namely the Composite Stock Price Index (JCI), Philippine Stock Exchange (PSEi), Strait Time Index (STI), Kuala Lumpur Composite Index (KLCI), and Thailand Stock Exchange Index (SETI). Stock Price Index data was divided into in-sample data for Generalized Space-Time Autoregressive (GSTAR) modelling and out-sample data used to validate presumptive results. In-sample data was taken from January 4, 2021, to December 29, 2023, and then out-sample data for presumptive was as many as 5 from January 2, 2024, to January 8, 2024. From the modeling results, it was found that the mean MAPE value of the GSTAR model was smaller than that of the ARIMA model. Moreover, based on the presumptive results for the following 5 periods using the GSTAR (2.1) I(1) model, a Mean Absolute Percentage Error (MAPE) of less than 10% in each location. The values shows that GSTAR model is more accurate than the ARIMA model.

Keywords

• gstar
• arima
• stock price index