Comparing Prediction Methods of Artificial Neural Networks in Extracting Financial Cycles of Tehran Stock Exchange based on Markov Switching and Ant Colony Algorithm

摘要: The stock exchange is considered to be an important establishment finance long term projects, on one hand, and collect savings of private section. can a safe secure place invest surplus funds purchase corporate stocks. As recession prosperity in this market have great role stockholders` decision-making, it becomes vital predict these cycles. In paper, using model MSMH(4)AR(2), we extract the financial cycles market. Then, ant colony algorithm, determine most significant predictors neural networks. results show that PNN performs better predicting future with respect criteria mean squared error, root accuracy kappa coefficient.