The increasing emphasis on sustainable development in the real estate sector calls for decision-support methods that jointly account for financial performance, ESG risk, and evolving regulatory constraints. However, existing ESG assessment approaches often rely on static indicators or qualitative evaluations, limiting their ability to capture dynamic interactions among market conditions, regulation, and long-term sustainability. In this work, we present an integrated ESG-oriented decision-support framework for real estate investment that combines multiple existing modeling components. The approach integrates (i) a risk-aware profit optimisation module with portfolio-style objectives and resource constraints, (ii) an adaptive regulatory compliance module that models evolving ESG and regulatory requirements through time-varying penalties, and (iii) a risk-aware adaptive selection strategy (AREIS) based on reinforcement-learning-style sequential decision-making. The contribution lies in the principled combination and instantiation of these components for ESG-driven real estate investment, rather than in proposing new learning primitives. The framework further integrates heterogeneous data sources, including project-level attributes, geospatial context, and environmental exposure indicators, to construct interpretable ESG risk signals. Experimental results under realistic temporal splits show improved ESG risk prediction accuracy and higher ESG-adjusted investment utility compared to strong predictive and reinforcement-learning baselines, while reducing regulatory violations.
Xiao et al. (Tue,) studied this question.