Abstract
The rapid digital transformation of workplaces requires a comprehensive understanding of its implications for employee well-being, particularly regarding work-related factors and digital work challenges. Using data from the 2021 European Working Conditions Telephone Survey (EWCTS) across 12 European countries, we propose a novel methodological framework based on Random Forests (RF), specifically designed for class imbalance scenarios. The framework integrates multiple imputation for missing values, a density-based classifier, feature selection and interaction analysis based on tree topology, and optimized visualization tools. Our methodology combines supervised and unsupervised RF-based methods to identify key work-related factors influencing employee well-being and their relationships with digital work aspects. The results highlight psychosocial factors, particularly exhaustion, as predominant determinants of well-being, with recognition and engagement serving as mitigating influences. Feature interaction analysis underscores the critical role of telework intensity, revealing links to both negative outcomes (emotional exhaustion) and positive dimensions (work-life balance), demonstrating the complex relationship between well-being and digital work. Furthermore, a significant association exists between work-life balance and telework intensity, emphasizing the importance of time management in employees’ lives. This study advances both methodology, by introducing a novel RF-based approach for class imbalance problems, and theory, by identifying key work-related factors affecting employee well-being and their connection to digital work characteristics.
Cite as:
Demaria, F., and Cavicchioli, M. (2025). Employees’ well-being, work-related factors, and digitalization: a decision-tree approach for imbalanced data. Annals of Operations Research. doi: 10.1007/s10479-025-06702-9