Bankruptcy Prediction through Artificial Intelligence

Abstract

Bankruptcy prediction or corporate failure is considered a classic issue in both, academic and business communities. Bankruptcy risk is one of the most important factors (if not the most important one) to be considered when credit requests are screened or even existing debtors are evaluated. On the other hand, all potential stakeholders (shareholders, suppliers, customers, employees, creditors, auditors, etc.) have potential interest to identify if a company is on a trajectory that is tending towards failure. Commercial banks, public accounting firms and other institutional entities (e.g., bond rating agencies) appear to be the primary beneficiaries of accurate bankruptcy prediction, since they can use research results to minimize exposure to potential client failures. In addition to avoiding potentially troubled obligors, the research can also benefit in other ways. It can help in accurately assessing the credit risk of bank loan portfolios. Credit risk has been the subject of much research activity, since the regulators are acknowledging the need and are urging the banks to assess the credit risk in their portfolios. Measuring the credit risk accurately also allows banks to engineer future lending transactions, so as to achieve targeted return/risk characteristics. The other benefit of the prediction of bankruptcies is for accounting firms. If an accounting firm audits a potentially troubled firm, and misses giving a warning signal then it faces costly lawsuits (Atiya, 2001). A series of techniques have been applied in literature. Econometric / statistical methods have first appeared in literature: In late 1960’s (multiple) discriminant analysis (DA) was the dominant method; during the 1980’s logistic analysis. In the 1990’s artificial intelligence starts appearing in financial literature with neural networks (Odom & Sharda 1990) serving as an alternative to statistical methods demonstrating promising results. The goal of this chapter is therefore two-fold: First, it intends to give an overview of the artificial intelligence techniques successfully applied to the problem, ranging from the first neural network applications to recent applications of biologically inspired algorithms, such as genetic algorithms. Then, two kernel based methods, namely the Radial Basis Function Neural Networks and the Support Vector Machines are applied to the bankruptcy problem.