Prediction of bankruptcy on industry classification

Abstract

This study predicts bankruptcy by making predictions after separating the entire industry into individual industries via the decision tree method and support vector machines (SVMs), and then it compares two method, decision tree and SVM. This research predicts bankruptcy in two steps. The first step is to predict using decision tree method and the second step is to predict using support vector machine for 11 industries. The financial statements of the listed companies in the Tokyo Stock Exchange, the Osaka Securities Exchange, and other stock exchanges were used as data in this study. The data of 244 bankrupt companies that went bankrupt between 1991 and 2014 are used. On the other hand, data of 64708 non-bankrupt companies that did not go bankrupt between 1991 and 2014 for 24 years are used. The data is acquired from Nikkei NEEDS database. In the decision tree, the multi-step branching process is stratified, and the bankruptcy prediction is performed in the tree diagram. In SVM, prediction of bankruptcy is almost perfectly conducted to discriminate the companies for each industry.