Wybrane metody klasyfikacyjne oraz ich efektywność w prognozowaniu upadłości firm

Classification methods are recognised as useful tool for bankruptcy prediction. Among them the most popular are: linear discriminant function, Logit model, neural network and classification tree. The ideas and basic formulas of these methods are presented in the paper. Some examples of application those procedures, which were published in world and Polish literature, are mentioned in the following parts of the paper. Some effectiveness conditions of presented methods are discussed. In the conclusion it has been stressed, that the precision of bankruptcy prediction not strictly depend on classification method which has been used. Sources of errors in bankruptcy prediction has been discussed on the end of the paper. Among them important are: valuated character of financial ratios, as an impute variables in such models, problems in samples selection, which usually hasn't random character and unstable character of considered populations. Probability of firms' collapse strongly depends on the stage of business cycle.