Comparing the Accuracy of Markov Switching – AR and Prophet Models in Predicting the Blue Bird Stock Prices

One form of investment asset that is in high demand for profit is stocks. However, stock prices fluctuate, so a mathematical model is needed to model the movement and calculate stock price predictions. Stock price movements often form several groups (states) of change, so the Markov Switching Autoregressive (MS-AR) model can be used to model and calculate stock price predictions. In addition, stock price movements often contain trend and seasonal patterns, so the Prophet model can be used to model movements and calculate stock price predictions. In this study, the Prophet model is modified by generating random numbers that spread normally with parameter values obtained from the error value of the Prophet base model. This study aims to compare the performance of the MS-AR model with the Prophet model in predicting BIRD stock prices. This research is a quantitative study with secondary data in the form of BIRD stock closing price data for the period 11 February 2023 to 11 February 2024. In this study, two models, MS-AR and Prophet, were built separately. In the MS-AR model, it is necessary to pay attention to the assumptions of the data used, namely normal distribution and stationary. In the Prophet model, there are no special assumptions like those of the MS-AR model, but the Prophet model is good for data containing trends and seasonal patterns. The results of this study show that among the MS-AR models, the MS(2)-AR(3) model is the best model. In addition, the results show that the modified Prophet model performs better than the basic Prophet model. The goodness of model performance is measured by the Mean Absolute Percentage Error (MAPE) metric, with MAPE values for each model being 5.54% for MS(2)-AR(3), 3.38% for the Prophet base model, and 2.88% for Prophet modification. Based on the MAPE value, the Prophet (modified) model is able to predict the closing price of shares better than the MS(2)-AR(3) and Prophet (basic) models. The results of this study can be used by investors as a measuring tool in reading and determining stock price predictions.