Abstract Climate change endangers the Carpathian region by increasing the risk of fires. In response, our study provides a harmonised dataset with twenty-seven variables and develops an interpretable machine learning-based framework for assessing fire susceptibility across all seven countries of the region. We applied a two-stage process: first, using various feature selection techniques to refine predictors before the modeling phase, and second, utilising the SHAP framework to interpret model predictions. Between these steps, advanced machine learning models were optimised and trained in the H2O environment, demonstrating high predictive accuracy. Our findings revealed eight fire susceptibility clusters. The resulting dataset, susceptibility maps, and detailed interpretative insights serve as a valuable resource for local communities and policy-makers in the region.