Coherent forecasts for tourism demand with automated immutability constraints

This study tackles key challenges in tourism demand forecasting within a hierarchical time series framework. To ensure coherence across aggregation levels and improve forecasting performance, we incorporate immutability constraints that preserve forecasts for strategically important nodes. Two automated selection methods are proposed to identify such nodes: (i) a clustering-based approach that ensures dispersion across levels, and (ii) a penalized optimization approach that selects immutable nodes based on data-driven criteria. Through Monte Carlo simulations, and two empirical applications, we demonstrate that the proposed methods improve forecast accuracy, robustness and flexibility while preserving interpretability. The framework is model-agnostic with respect to base forecasts and provides tourism managers with a scalable, data-driven tool to focus on critical segments, improve resource allocation, and support strategic planning in tourism management.