Enhancing Bankruptcy Prediction Using Deep Learning Techniques

Abstract

Forecasting corporate bankruptcy remains one of the most important issues in the financial and insurance sectors, where predictions can make a substantial difference between the solvency and bankruptcy of companies, and therefore between economic stability and instability. While machine learning already plays an important role in finance, this study specifically explores the power of deep neural networks to predict business bankruptcy. Our objective is to construct a model that will help stakeholders identify business failure signs as early as possible. It uses a multilayer DNN to analyze five years' financial ratios extracted from the balance sheet and income statement of public companies. Its performance was evaluated against established measures such as accuracy, precision, recall, and the F1 score. The results show that the proposed deep neural network has a high accuracy for predicting bankruptcy, demonstrating its strength as a predictive model. We prove that DNNs are strong and feasible for this task. Based on the findings, we suggest that the next research direction should aim at developing models to attain standardization and predict bankruptcy for private companies in emerging markets.