Bankruptcy Prediction Model in Tehran Stock Exhange

Bankruptcy is the last part of the companies’ life cycle and it affects a wide range of financial participants. Bankruptcy prediction is important not only for financial participants but also for politicians, lawyers and researchers.As legal process of declearing bankruptcy is a time consuming and beurocratic issue, in this paper, we use the concept of bankruptcy triggering asset value to detect bankrupt companies in Tehran Stock Exchange. We use neural network as a powerfull tool for predicting bankruptcy. For training purposes, firstly back propagation method was used and then the aim became to improve the prediction results by using genetic algorithm and particle swarm optimization. Our findings illustrated that genetic algorithm acts better than back propagation method but we do not have enough evidence to prove that generaly particle swarm optimization acts better than genetic algorithm.We also comared financial ratios’ power versus market data for predicting bankruptcy. To do this, we used three groups of data, financial ratios, market data and simultaneously using both financial ratios and market data. It was discovered that market data is a better mean for predicting bankruptcy. Our findings show that using particle swarm optimization for training method and market data as an input for predicting bankruptcy could lead to 92.6% correct prediction of bankruptcy in test sample.