PERFORMANCE ANALYSIS OF GRADIENT BOOSTING MODELS VARIANTS IN PREDICTING THE DIRECTION OF STOCK CLOSING PRICES ON THE INDONESIA STOCK EXCHANGE

Abstract

Accurately predicting stock market trends remains a significant challenge for investors due to its dynamic nature. This study explores the performance of Gradient Boosting models, including XGBoost, XGBoost Random Forest, CatBoost, and Gradient Boosting Scikit-Learn, in predicting stock market trends such as sideways movement, uptrends, downtrends, and volatility. Using four datasets from the Indonesia Stock Exchange, the research integrates technical, fundamental, and sentiment data, encompassing 37 features. Modeling and testing are conducted using Orange tools and Python, with performance evaluated through metrics such as Mean Absolute Percentage Error (MAPE), R-squared (R²), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). Results indicate that XGBoost and XGBoost Random Forest consistently outperform other models in predicting stock price movements. These findings highlight the potential of Gradient Boosting models in providing accurate and reliable predictions, offering valuable insights for investors, financial analysts, and researchers to enhance investment strategies and adapt to market fluctuations effectively.

Keywords

• catboost
• gradient boosting
• performance analysis
• random forest
• stock prediction
• xgboost