Research on household energy poverty is crucial for addressing basic human needs, highlighting the importance of conducting studies across different countries. This study examines energy poverty in the Visegrad Group countries using 2022 data from the EU Statistics on Income and Living Conditions survey, with the ‘inability to keep the home warm’ indicator as a measure. The study aims to identify socio-economic factors influencing energy poverty and examine how their impacts differ across the examined countries. Logistic regression, combined with machine learning techniques, was applied for this purpose. Furthermore, the study evaluates the classification results of logistic regression and three machine learning methods—CatBoost, Balanced Random Forests, and Extreme Gradient Boosting—on imbalanced data. It was found that, among the three machine learning methods used, Balanced Random Forests performed the weakest. Logistic regression, effective for our imbalanced data, complements the results and provides deeper insights into the socio-economic factors influencing energy poverty. The study found that Slovakia had the highest percentage of households vulnerable to energy poverty, while Czechia had the lowest. Income, household type, and the presence of disabled individuals were found to be important across all countries. However, other factors varied in their influence from one country to another, highlighting the need for country-specific analyses. Monitoring households’ exposure to energy poverty is a challenge for future social policy and the use of different methods provides an in-depth view of this complex issue.

One of the goals of macroeconomic analysis is to rank and segment enterprises described by many financial indicators. The segmentation can be used for investment strategies or risk evaluation. The aim of this research was to distinguish groups of similar objects and visualize the results in a low dimensional space. In order to obtain clusters of similar objects, the authors applied a DEA BCC model and archetypal analysis for a set of companies described by financial indicators and listed on the Warsaw Stock Exchange. The authors showed that both methods give consistent results. To get a better insight into the data structure as well as a visualization of the similarities between objects, the authors used a new approach called the PHATE algorithm. It allowed the results of DEA and archetypal analysis to be visualized in a low dimensional space.