Hybrid ARIMA-GARCH Model with Walk-Forward Method On LQ45 Stock Price Forecasting

Abstract

Stock investment offers returns but also risks, such as potential capital losses due to declining stock prices. To mitigate these risks, investors use forecasting models, and one common approach is time series forecasting. The ARIMA model captures linear patterns in data, while the GARCH model handles time-varying volatility. This study uses a hybrid ARIMA-GARCH model with the Walk-Forward method to predict the daily closing prices of LQ45 index stocks from January 2022 to May 2024, utilizing data from Yahoo Finance. The Walk-Forward approach divides the data into 80% training and 20% testing, ensuring the model is tested on unseen data for more realistic evaluation. The process includes fitting the ARIMA model to stock return data, testing for heteroscedasticity, and building the hybrid ARIMA-GARCH model. The best model, ARIMA(1,0,0) – GARCH(1,1), was selected based on the lowest AIC value of -3004.88 for ARIMA and -6.83 (AIC) and -6.78 (BIC) for GARCH. This research contributes to stock forecasting by applying high-frequency data and the Walk-Forward validation method, offering a more accurate assessment of the model’s performance. It also enriches time series analysis methodology in the Indonesian stock market by combining ARIMA and GARCH models, optimizing model parameters using AIC and BIC criteria for stock price prediction.