Prediction of Corporate Bankruptcy using Financial Ratios and News

Abstract

A corporate’s insolvency can have catastrophic effects on not only the corporate but also on the returns of its lenders and investors. Predicting bankruptcy has been one of the most sought-after areas for researchers for many decades. This study involves predicting the bankruptcy of the United States corporates using financial ratios and news data. The financial ratios of the companies were extracted from yearly financial reports of the companies, and the news data of the companies was scrapped from online newspapers, reports and articles using Google News. The news data was analyzed for negative and positive sentiments. The sentiment scores, along with the financial ratios of the companies, were given as features to the machine learning models. Various models were analyzed for their results such as Random Forest, Logistic Regression and Support Vector Machines (SVM). The study finds the best results from the random forest model with an accuracy of 90%. Moreover, the significant feature importance of the sentiment score given by the model proves that unstructured data, such as news, can play a crucial part in predicting bankruptcy in conjunction with the structured data, such as financial ratios.