Utilisation de variables non financières dans le cas de la prédiction de faillite d'antreprises de moins de 5 ans : une approche multicritère pour le cas belge

Business failure prediction is a topic of outmost importance for corporate shareholders, bank managers, investors, suppliers, government officers, etc. Due the significance of business failures, many prediction models have been developed. Statistical and econometric procedures (multiple discriminant analysis, logit or probit) have been among the most widely used methods in this field. Other methods have also been investigated: mathematical programming, multicriteria decision aid, machine learning (neural networks, genetic algorithms, decision trees), etc. Most of the business failure prediction models developed in the past have been based solely on the use of financial ratios constructed from published accounting data, which are generally easy to obtain. In Belgium, however, corporate financial data are not always available. This situation clearly highlights the necessity of using non-financial information to predict bankruptcy