Performance Accuracy of Modified Hybrid Geometric Brownian Motion Kalman Filter and Markov Switching Methods in Indonesian Stock Prices Forecasting

This study develops and evaluates two hybrid stochastic models, Hybrid Geometric Brownian Motion-Kalman Filter (GBM-KF) and Hybrid Geometric Brownian Motion-Markov Switching (GBM-MS), in stock price forecasting in the Indonesian capital market. The hybrid GBM-KF model utilises the dynamic filtering mechanism of the Kalman Filter to adaptively update parameter estimates in response to price fluctuations, while the hybrid GBM-MS combines regime switching dynamics based on Markov processes to represent changes in market conditions between stable and volatile phases. Using daily closing price data for BBCA and BBRI shares for the 2024 period, the performance of the models was tested using Mean Absolute Percentage Error (MAPE). The results show that both models achieved best fitting conditions without indications of overfitting or underfitting, with test MAPE values of 4.30% (GBM-KF) and 1.38% (GBM-MS) for BBCA, and 8.91% (GBM-KF) and 1.60% (GBM-MS) for BBRI. Overall, GBM-MS showed the highest accuracy and the most effective ability to capture changes in market structure and stock price regime dynamics in Indonesia.