The Application of Machine Learning Models in Company Bankruptcy Prediction

Abstract

With the economy increasing, the number of enterprises increases gradually, and it is accompanied by the growth in quantity of bankruptcies.Therefore, predict bankruptcy is becoming more andmore important. It could not only help make the correct decision, but also reduce losses. There are several traditional methods are commonly used to predict the corporate bankruptcy conditions. But the traditional methods for bankruptcy prediction are mainly based on human subjective judgment and lack quantitative analysis. So it is for traditional methods to compete with the new and advanced Machine learning algorithms. Machine learning algorithms explore patterns based on objective data analysis, which develop rapidly and have strong learning ability. So we're going to apply machine learning to the prediction of corporate bankruptcy. We use data on the bankruptcy situation of Polish companies in 2007-2013 and construct a model by SVM and random forest algorithm separately. And then, we further use weighted methods to solve the problem of sample imbalance. According to the research, Random forest performs better than SVM in company bankruptcy prediction with accuracy higher than 70% in different years.