Porównanie skutecz-ności klasyfikacyjnej wybranych metod prognozowania bankructwa przedsiębiorstw przy losowym i nielosowym doborze prób
Comparison of classification accuracy of selected bankruptcy prediction methods in the case of random and non-random sampling technique

Summary/Abstract: One of the sources of errors being committed in the process of bankruptcy prediction is a method for selecting samples. During the construction of a sample of balanced structure, the most popular non-random approach is based on pairing up bankrupt companies with non-bankrupt ones. The alternative to pair-matched sampling is simple random sampling with replacement. The article presents a comparative study of selected failure prediction models, taking into account the random and non-random technique of samples selection. Data was divided into a training group and a testing group in a ratio of both 7:3 and 6:4. A modification of Logit model consisting in intro-ducing a time factor into a model is also presented. The usefulness of this modification in the context of its classification accuracy improvement for two aforementioned techniques of sampling was verified.

• Details
• Contents