<p>Corporate bankruptcies are an integral part of the functioning of economies. Researchers are increasingly using statistical methods to predict company bankruptcies. However, this raises the problem of choosing the approach, method, procedure or algorithm that is most effective from a forecasting perspective. This paper presents the results of applying the author’s proposal of a new research approach based on selected statistical learning methods to predict corporate bankruptcy. We formulated a research question: whether combining unsupervised and supervised statistical learning methods can improve the accuracy of corporate bankruptcy forecasting. We have conducted a scoping review of the literature based on Scopus database resources. In this way we identified a research gap in the literature. The gap lies in the lack of works in which researchers used unsupervised statistical learning methods to generate variables that inform about the class membership of objects, and such variables incorporated then into the set of predictors in classification methods. The presented article addresses this research gap. In our opinion, the research results presented in the paper partially fill this research gap. The study used financial data from industrial processing companies operating in Poland. We forecasted the bankruptcy of companies one year in advance. The study’s results confirmed that some combinations of statistical methods used in unsupervised and supervised machine learning, which were developed and applied in the analysis, improve the accuracy of predicting corporate bankruptcy. The added value of the presented work is partially to fill the research gap in corporate bankruptcy prediction identified in the scoping literature review.</p>