Building site selection and protection: an integrated decision-making model based on architecture of a geographic information system and comprehensive building conservation index

Background Frequent natural disasters make building site selection and protection considerably challenging because they require comprehensive multi-hazard risk assessments that must account for the complex interactions between different disaster types, each with distinct and often conflicting site requirements. This complexity is further exacerbated by climate change, which increases the frequency and intensity of extreme weather events, rendering traditional risk assessment methods based on historical data increasingly inadequate and necessitating more sophisticated, forward-looking approaches to ensure long-term building resilience and safety. Moreover, the existing mechanisms for disaster insurance decision-making and resource allocation often lack efficiency. This study aims to address the inefficiency of the disaster insurance decision-making and resource allocation mechanism by proposing an innovative optimization model that facilitates the development of building site selection and protection strategies, especially in disaster-prone areas. Methods The proposed model assimilates economic-loss data, leverages the Holt-Winters time-series approach to sharpen decisions regarding disaster insurance, and deploys Random Forest to derive the relative weights of critical risk indicators. The value of the model is demonstrated using Mexico City as a case study. Specifically, the study employs the Architecture of a Geographic Information System (ArcGIS) Site Suitability Model to assess land development potential and applies K-means clustering to identify areas with the highest development potential. Furthermore, a Comprehensive Building Conservation Index is developed using data on six high-value buildings, with the Museum of Anthropology emerging as a primary conservation target. Finally, a costed conservation plan encompassing both short- and long-term phases is proposed. The data/code is available at https://github.com/Luyu-star/Building-site-selection-and-protection.git . Results The findings demonstrate that the optimization strategy proposed in this study provides valuable decision-making support for disaster response, risk management, and urban planning in Mexico City. The integration of machine learning and GIS technologies offers a robust and efficient strategy to tackle these aforementioned challenges. The Holt–Winters forecasting component achieved R² values of 0.739 (95% CI [0.712–0.765]) for Armenia and 0.642 (95% CI [0.601–0.683]) for Mexico. Overall, the proposed model offers a more integrated and coherent decision-making framework compared to conventional fragmented approaches.