Bankruptcy Prediction and Credit Scoring: A Review of Recent Developments Based on Hybrid Systems and Some Related Patents

Javier      de Andres; Pedro      Lorca; Fernando      Sanchez-Lasheras; Francisco      Javier De Cos-Juez

doi:10.2174/2213275911205010011

Abstract

This paper reviews some recent developments from the field of Artificial Intelligence aimed at the determination of credit scores of firms and individuals. Specifically, the academic research efforts which are reviewed are those that consist of hybrid systems. These systems combine two or more intelligent techniques in several forms to derive the advantages of all of them. There are four ways to elaborate a hybrid system: first, by constructing a hybrid algorithm which tightly combines two or more single algorithms. Second, by using an ensemble, which is the combination of the results of a pool of models. Third, by using an algorithm to select the relevant variables and another one to perform the scoring. And finally, by clustering individuals and using this information for the scoring stage. Some patents have used these developments, but in our opinion only a small part of the research has been translated into patents and practical applications.

Keywords: Artificial intelligence, bankruptcy, classification, clustering, credit applications, credit risk, credit scoring, ensemble classifiers, financial distress, financial ratios, financial statements, hybrid systems, feature selection, hybrid models, solvency, statistical learning