<![CDATA[This study investigates the efficacy of Prophet, XGBoost, and Ridge Regression in forecasting stock prices of four major Indonesian banks—Bank Central Asia (BBCA.JK), Bank Negara Indonesia (BBNI.JK), Bank Rakyat Indonesia (BBRI.JK), and Bank Mandiri (BMRI.JK)—using daily historical data from January 2020 to March 2025, sourced from Yahoo Finance. The banking sector's volatility, evidenced by year-to-date declines ranging from 7.59% (BBCA) to 22.69% (BMRI) as of May 1, 2025, underscores the need for accurate predictive models. Performance was evaluated using Mean Absolute Error (MAE) and Root Mean Square Error (RMSE), revealing Ridge Regression as the superior method, consistently achieving the lowest errors (i.e., MAE of 23.81 for BBNI.JK and RMSE of 55.75 for BBCA.JK). Prophet exhibited the highest errors, suggesting its seasonal focus is less suited to stock price unpredictability, while XGBoost performed moderately better but lagged behind Ridge Regression. The results highlight Ridge Regression’s effectiveness in handling multicollinearity and noise in financial data. Our discussions emphasize the importance of model selection based on data characteristics, with implications for investment decision-making in the Indonesian market. This research contributes to the field of computational finance by providing a comparative analysis that not only identifies Ridge Regression as a superior method for forecasting stock prices but also illuminates the limitations of popular models like Prophet and XGBoost in handling financial data's unique characteristics, guiding future model selection and development. Future research could explore hybrid models to enhance accuracy across varied market conditions, addressing the study’s 60-day forecasting horizon limitation.]]>
Copyrights © 2025
