Implementation of Hidden Markov Model (HMM) to Predict Financial Market Regime

This work performs how to implement the concept of Hidden Markov Model (HMM) to find financial market trend for given only the observed state obtained from the stock price. The considered market trend is set as a hidden state, that in the financial technical analysis known as Bearish, Bullish, and Sideway, which are important for decision making of stock trading in order to recognize the good moment to sell, to buy or to just hold the shares. In order to obtain the most likely sequence of hidden states through HMM, which is computationally can be a dynamic programming problem, we explain how the Viterbi algorithm work for the case in this study. To get the stock price prediction as observation states, the ARIMA model is used based on experimental trial of fitting model, then use the result as a predicted observed states that be the input to predict the market trend using HMM for the short period of future time. Several interesting results of hidden market trend and its study are also provided, including the accuracy, precision, recall and the consistency of the model to the given data set.