Iran Stock Market Prediction Based on Bayesian Networks and Hidden Markov Models

Stock market behavior is one of the most complex mechanisms, considered by researchers. Financial markets are influenced by the external and internal factors. External factors such as political and social factors are not measurable, so prediction the trend of stock markets is focused on internal factors.
This study suggests a hybrid approach based on Bayesian Networks and Hidden Markov Models to predict trend of stock market. The used variables are 6 index of Tehran Stock Exchange, which have the most correlation coefficient with target stock, and 22 technical indicators. Bayesian networks are utilized to find the relationships between variables, and the effect of each variable in prediction considered from conditional probability tables. Hidden Markov Model is designed for sets of extract from Bayesian networks.
The proposed model tested on four company’s stock names Mobarakeh Steel, Iran Khodro, Mellat Bank and Iran drug.
The average accuracy of the proposed system is 83.26 %. The experimental results show that the suggested procedure has higher performance for prediction of stock markets in comparison with other previous methods.