Bankruptcy prediction: literature survey of the last ten years
Abstract
Bankruptcy is an important topic for a number of people (shareholders, banks, investors, suppliers,...). For this reason a lot of models were developed in orderto predict it. Statistical procedures (multiple discriminant analysis, logit or probit) were among the most used methods in this kind of problem. However, parametric statistical methods require the data to have a specific distribution. In addition to the restriction on the distribution involved, multicollinearity and autocorrelation could lead to problems with the estimated model when statistical methods are used. Because of these drawbacks, others methods have been investigated : multicriteria methods (UTA,Electretri,...) or machine learning methods (i.e. neural networks, genetic algorithms, decision trees, rough sets,...). Our main target is to provide a survey of the literature of the last ten years but also to have a larger view than usually by evoking causes, symptoms and remedies of bankruptcy.