Identifying stock market regimes in Indonesia's non-primary consumer goods sector using a Gaussian Hidden Markov Model

Stock price movements in Indonesia's non-primary consumer goods sector reflect a complex interaction between market sentiment and macroeconomic fluctuations. This study identifies latent market phases and evaluates price forecasting for three selected stocks—MNCN, FAST, and ACES—which span different sub-industries and liquidity profiles, using a Gaussian Hidden Markov Model (GHMM) calibrated on daily open, high, low, and close (OHLC) prices over 620 trading days (January 2024–August 2026). Model selection via the Bayesian Information Criterion, assessed through re-estimation with multiple random seeds to examine convergence stability, identifies five latent states for MNCN and six for both FAST and ACES, indicating that regime structure varies across issuers even within the same sector. A likelihood-based forecasting procedure is then evaluated over an extended out-of-sample period of 124 trading days (February–August 2026), yielding mean absolute percentage error (MAPE) values of 1.79% for MNCN, 4.65% for FAST, and 6.22% for ACES. However, the GHMM forecasts did not outperform a random-walk benchmark for any of the three stocks. The findings therefore highlight the GHMM's value in characterising issuer-specific market regimes and monitoring risk, rather than improving short-horizon point forecasts.