COMPARATIVE ANALYSIS OF HIDDEN MARKOV MODELS AND ARIMA FOR STOCK PRICE PREDICTION: A PERFORMANCE EVALUATION

Abstract

Stock market prediction involves using analytical techniques to predict stock prices or indices based on records. This makes accuracy essential to the depositors, dealers, and monetary organizations operating near marks regarding knowledgeable buying and selling, along with effective management of the portfolios. Strong influence would be exercised in effective models to predict changing investment strategies in volatile markets to enhance risk management. Reliability in forecasts would give an advantage over the competitors as it helps stakeholders know the market's movement and optimize returns, thereby minimizing losses. This paper compares the performances of two popular models used to predict stock price: Hidden Markov Models (HMM) and Auto-Regressive Integrated Moving Average (ARIMA). Our three key metrics under consideration in this analysis are MAPE, SMAPE, and Coefficient of Determination, which is R². The outcomes indicate that HMM dramatically outperforms ARIMA under all evaluations. For instance, HMM attains a value of 1.51 MAPE, 1.51 SMAPE, and an R² value of 0.8882. This depicts great precision with a very good fit for the model. On the contrary, ARIMA shows a much greater MAPE value of 5.91 and an SMAPE of 5.71 and holds an even terrible R² at -0.3709 thus indicating that it is quite a bad fit. The analysis also reveals a good performance in time series for HMM concerning financial domains that include stock price predictions where they require accurate forecasts and reliable information to determine what decisions to be taken. An improved concept and robust framework on the implementation of HMM into financial forecasting research has been given for accuracy that surpasses traditional models ARIMA.