Stock Price Index Prediction of Several ASEAN Countries Using Panel Data Regression and Machine Learning Algorithm

The analysis of major stock indices plays a crucial role in understanding market sentiment and predicting economic performance in the ASEAN region. Several studies have examined panel data regression and machine learning methods to analyze macroeconomic variables related to stock indices, including boosting models such as XGBoost, CatBoost, and LightGBM, which have proven effective in handling the complexity of multivariable financial data. This study aims to explore the use of the Fixed Effect Model (FEM) to measure the impact of variables such as gold prices, dollar exchange rates, bond yields, and commodity prices on major stock indices, as well as employing machine learning for short-term forecasting. Daily data from 2019-2024 for ASEAN countries was used, with models selected based on significance tests, accuracy, MAE, MSE, and R2. The results show that LightGBM achieves the highest accuracy, with an MAE of 31.05 and 53.77, and an MSE of 3105.61 and 9786.65 on training and testing data, respectively, achieving 100% R2 on both. CatBoost and XGBoost also perform well, with high accuracy and low error values. Results show LightGBM as a reliable tool for projecting ASEAN stock market trends.