A Data-Driven Information System Architecture for Analysis of Environmental, Geopolitical, and Health Risks in the EU-27

The increasing interdependence between environmental degradation, geopolitical instability, and public-health pressures requires structured information-system architectures capable of integrating heterogeneous data and transforming them into decision-support knowledge. In this context, this study develops a data-driven information system architecture for the exploratory analysis of environmental, geopolitical, and health-related risks in the EU-27 during 2013–2023. The proposed system is structured as a multi-layered analytical pipeline designed to process country-year panel data and generate interpretable outputs. The methodological framework integrates Principal Component Analysis (PCA) for exploratory dimensionality reduction, K-Means clustering for structural pattern identification, a RandomTree classification model for translating cluster membership into decision rules, and a Two-Part Fixed Effects Model. Experimental results indicate an optimal and interpretable clustering configuration at k = 3, revealing three broad structural profiles among EU Member States. A moderate positive relationship is identified between greenhouse gas emissions per capita (GHGE) and health expenditure (SHA) (r = 0.34), while geopolitical risk (GPR) exhibits weak and statistically insignificant associations. This association is interpreted cautiously, as it may reflect the combined effect of industrial activity, environmental exposure, economic development, and the higher financial capacity of some Member States to allocate resources to healthcare systems. The results indicate the dominant contribution of GHGE and SHA in differentiating the identified profiles, while GPR shows limited explanatory power within the analyzed context. The RandomTree model achieved an accuracy of 93.58% in reproducing the cluster labels; however, it is used as an interpretability layer rather than as an independent validation of clustering. The system supports the identification of vulnerability-related structural patterns and provides an exploratory basis for future data-driven monitoring and early-warning applications.