Integrating Social Actor Presence into an Explainable Hybrid Workflow for Urban Heritage Conservation: The Bellavista Case in Tomé, Chile

Social actors play a key role in urban heritage conservation by shaping decisions, conflicts, and territorial transformations. However, their incorporation into spatial modelling remains limited. This study evaluates whether spatialised social actor presence provides additional explanatory and predictive information for land-use change modelling in Bellavista, Tomé, Chile. The workflow combines Least Absolute Shrinkage and Selection Operator (LASSO)-based variable selection followed by an unpenalised binomial Generalised Linear Model (GLM) for interpretation, Random Forest for prediction, and Cellular Automata for spatial simulation, with predictive performance assessed through spatial cross-validation. The contribution of the social dimension was evaluated by comparing models with and without social actor presence variables under spatial cross-validation. The results show modest but consistent improvements in spatial predictive performance when social variables are incorporated. Social actor presence showed transition-specific effects: Companies and Academia were significantly associated with the Historic-to-New transition, whereas no individual social variable was significant for the Historic Open Space-to-New transition. Nevertheless, incorporating social variables consistently improved Random Forest performance under spatial cross-validation in both transitions. The retrospective simulation reproduced the main observed patterns of change, while the 2030 scenario indicates a substantial reduction in historic open spaces.