Stock market prediction with Hidden Markov Model

Today, Stock investment is an important part of the economy of the country. Therefore, forecasting changes in the behavior of market has become significantly important to shareholders. In the past years, classic methods often were used to forecast changes in the market behavior, but, in recent years, intelligent methods increasingly have been applied to forecast the behavior of stock market. Using intelligent methods, the accuracy and precision of forecasts have increased to some extent, but further researches still can be carried out to improve the precision of forecasts. The aim of the present research was to develop an appropriate model for increasing the accuracy and precision of forecasting the behavior of Tehran Stock Exchange by using Hidden Markov Model. In this model, normal conditions were supposed for the market and it was also assumed that the changes in the market behavior were independent from factors such as political issues, war and natural disasters including flood, earthquake and fire. First, the model was trained by Baum-Welch Algorithm and after implementing the progressive forecast algorithm for three specific industries in MATLAB, it was found that, in comparison with the conventional methods, accuracy and precision of forecasting had been improved by 2 percent, on average. Moreover, by increasing the number of the days included in a specific interval considered for a certain industry, the precision of training algorithm increased and, consequently, the precision and accuracy of the algorithm increased, too.