Czy warto budować modele prognozowania upadłości przedsiębiorstw uwzględniające kryterium przynależności sektorowej?

Forecasting bankruptcy of enterprises is a very important area, and it is of interest to a wide group of stakeholders. For this reason, since the beginning of the last century, continuous work aimed at developing effective tools used to this end has been undertaken. The main goal of the study presented in this article is to verify the legitimacy of constructing industry models for forecasting the bankruptcy of enterprises. The research was conducted on the basis of finan-cial data of 800 Polish companies for the years 2015–2019, drawn from the EMIS Professional database. The analysis used six methods of variable selection, six machine learning methods and 500 distinct random training and testing samples. The obtained results lead to the conclusion that there are different determinants of bankruptcy in different sectors of the economy. Models constructed on the basis of a group of enterprises conducting homogeneous economic activity make it possible to achieve, on average, higher classification accuracy rates, which demonstrates that the construction of such models is justified.