You are here

Block title
Block content

Deal- to- Deal Marginal Efficiency Dynamics of Serial US Banking Acquirer



In this study, we apply several machine learning techniques including extreme gradient boosting (XGBoost), support vector machine and a deep neural network to predict bankruptcy using financial data of 3728 Belgian Small and Medium Enterprises (SME’s) during the period from 2002 to 2012. Xavier Bredart’s (2014) shallow neural networks model predicted bankruptcy of Belgian SMEs with about 80% accuracy using three features. Using the above-mentioned machine learning techniques, we predict bankruptcies with an overall accuracy of 82-83% using either 3 or 5 features. While the improvement of the overall accuracy of these models over the Bredart’s shallow network is modest 2%, there is 17% improvement in correctly classifying healthy firms. The limitation in the accuracy of the models stems from the inherent inseparability of the two categories of companies.
Key words: Bankruptcy, deep learning, support vector machine, extreme gradient boosting, Belgian SMEs
Note: To be presented in the 2019 Mid-Atlantic Region Meeting, Pittsburgh,2-4th of May.

Paper No.: