Application of Corporate Governance indices to predict Bankruptcy of Companies using Machine Learning

Bankruptcy Prediction of companies is becoming increasingly significant in recent times as it enables stakeholders to act quickly and reduce their financial losses. To create bankruptcy prediction models, many machine learning techniques have been applied using financial features. However, there have been comparatively less research on how non-financial features like Corporate Governance indices can be used to predict a company’s performance. Hence, this thesis is motivated by the need of further research within bankruptcy prediction influenced by Governance indices using traditional machine learning models and neural networks as there has been very less research using only the governance indices as the contributing features. These governance indices can be for example the age of the company, the number of shareholders, the number of board members etc of a company. In our thesis we aim to predict the bankruptcy of 160,000 Italian small and medium sized enterprises and understand the impact of Governance indices as important features related to bankruptcy prediction in companies. We try to understand which machine learning models predict the bankruptcy with maximum accuracy using only non-financial variables. As complex machine learning models are considered as a black box, we use Explainable Artificial intelligence to understand the most important governance indices that are contributing to the prediction of bankruptcy in each company as a greater number of stakeholders have started asking justification for the reason behind the prediction made by the models. We also discover the common patterns that is followed among all the companies in terms of the features contributing to the maximum accuracy.