Prediction of financial failure in business with machine learning methods and Altman Z-score

It is extremely important for the companies operating in a country to both maintain their own existence and the benefits they will provide to the country's economy. The globalization of the world economies and the resulting economic crises in the world negatively affect the economies of the states and the operating businesses. Within the framework of all these situations, it has become mandatory for businesses to be financially well-managed and to take the necessary measures before failure in order to prevent or be less affected by these crises. The aim of study is to compare machine learning methods and Altman Z-Score method by using 24 financial ratios the financial data of 178 companies in the manufacturing industry sector operating in Borsa Istanbul for the estimation of financial failure between the years 2015-2019, and to create financial failure prediction models up to 1, 2, 3, 4 and 5 years before financial failure. As a result of the comparative analysis, the Altman Z-Score of the companies operating in Borsa Istanbul was quite low until 5 years before the failure. Machine learning methods gave very good results according to Altman Z-Score. Among the machine learning models, the Random Forest method gave very good results up to 1, 2, 3, 4 and 5 years before the failure. Neural networks, support vector machines, and decision trees, respectively, nevertheless obtained very good prediction results according to the Altman Z-Score.