Stock Price Forecasting using Hidden Markov Models

Abstract

We address the challenging task of predicting stock values in this study, which has long been of interest to shareholders due to the complex, nonlinear, and dynamic nature of the stock market. The focus of this study is on forecasting future trends in stock market groups. The effectiveness of the Gaussian Hidden Markov Model (HMM) method for predicting stock prices is demonstrated, with evaluation conducted on four prominent groups: Apple Inc., CMCST corporation, Google LLC, and Qualcomm Inc. Historical stock market data, typically in the form of OHLC (Open-High-Low-Close) prices, is utilized as training data for the model. By leveraging this historical data, accurate predictions for the next day’s stock prices are made. The Mean Absolute Percentage Error (MAPE) metric is employed to evaluate the performance of the models, and the results are satisfactory. The proposed approach is applicable to predicting the stock prices of any company by training the HMM on the specific company’s stock dataset, making it a generalizable method.