A data-driven approach to spatial zoning and anomaly detection in the dynamic real estate network

We propose a novel statistical framework for dynamic zoning in real estate markets, leveraging spatial clustering and temporal change detection to ensure a robust and adaptive zoning mechanism. Our approach follows a two-stage procedure to systematically analyze zoning structures over time. First, we develop extended local indicators of spatial association algorithm to identify spatial clusters in a data-driven manner. These clusters are then integrated into a dynamic statistical network that incorporates geographical distance, capturing spatial dependencies and interactions between different regions. In the second stage of the algorithm, borrowing from the graph theory literature, we detect structural breaks in housing market dynamics using network Laplacians, which allow for identifying both gradual and abrupt shifts over time. This enables a precise understanding of how zoning structures evolve, distinguishing between short-term, long-term, and permanent shifts in spatial cohesion. Further, given the strong influence of zoning regulations on property pricing, we introduce a real-time property valuation score that captures market fluctuations while incorporating local spatial characteristics. Applying our methodology to weekly house price data from Greater London, we detect significant structural changes in market dynamics. Our proposed approach provides a statistically rigorous foundation for zoning policies, promoting data-driven decision-making in urban planning and real estate valuation.