Applying Bayesian network based methods to classify and select effective‎ ‎features in stocks

‎In this research‎, ‎four Bayesian network based classifiers are applied to predict stocks real return and risks‎. ‎In this method‎, ‎at first all possible features which can be effective on stocks risk and real return are identified‎. ‎In the next stage predictions are made by applying Na¨ıve ve Bayes‎, ‎tree augmented Na¨ıve Bayes‎, ‎conditional inference tree and general Bayesian networks‎. ‎To improve the prediction accuracies‎, ‎more effective features are chosen according to a feature subset selection method based on conditional inference trees‎. ‎The results show that the conditional inference tree is a proper tool for effective feature subset selection‎. ‎Also‎, ‎general Bayesian networks proved to be more efficient in classification and less sensitive to the choice of input features‎. ‎To illustrate the approach‎, ‎Tehran Stock Exchange (TSE) data sets from 2005 to 2014 is used‎. ‎These results argue that BN based classifiers deserve more attention in the data mining community‎.