Predicting the delisted companies of Tehran Stock Exchange using machine learning based algorithms

Delisted companies, despite their importance in the economic and social issues of society, is less considered in the financial literature. This issue is important because for each country, one of the criteria for economic measurement is the size of the capital market. Therefore, the delisted companies not only destroys the company's reputation, its stock price and the market for the sale of its shares, but also affects the growth of the market and the economy of each country. The present study seeks to review the financial statements and audit reports of active companies and compare it with delisted companies to design a model for forecasting delisted companies in the Tehran Stock Exchange with the help of artificial intelligence modeling techniques.

In this study, which was conducted on companies of the Tehran Stock Exchange, data related to three years before the delisting of 73 companies removed from the stock exchange from 2003 to 2019 in the first group and data of 148 active companies that are continuously. They were present in the stock market in the second group and were selected by systematic elimination method. Then, with data mining techniques, which are among the most efficient and up-to-date models of artificial intelligence, and with the help of multi-layered perceptron neural network classifiers, decision tree, and Bayesian theory classifiers, stock delisted companies have been predicted.

The findings show that the Bayesian classifier had the best performance and the multilayer perceptron neural network was in the second place and the decision tree classifier was in the third place.

Originality/Value:

 Little research has been done in the field of predicting delisted companies from the Iran capital market. This study by filling this gap, suggests to researchers to use other classifiers, combine several classifiers together to better cover the errors of each, combine classifiers with each other and weigh in a way that is more accurate, add other variables influential in the dismissal of companies, including the ownership structure and shareholder composition can have other results.