Optimization of Park Green-Space Site Selection in Changsha Based on Accessibility and Machine Learning

Fine-scale site selection of urban parks serves as a core measure to guarantee spatial equity of public service spaces for residents and advance the sustainable development of urban ecological spaces. Among relevant tasks, balancing the supply and demand of green spaces stands out as an essential foundation for maintaining long-term stability of urban human well-being and ecosystems. Green-space supply is defined as the stock of various existing urban parks within the city, while green-space demand is quantified via grids generated based on residential communities in Changsha. Existing research on urban park site selection lacks a full-process coupled framework, fails to accommodate differentiated layout demands for multi-level parks, and struggles to reconcile the sustainable operation and long-term ecological empowerment of urban green-space systems. Taking the main urban area of Changsha as the research scope, this study divides the study area into grid units to analyze the spatial differentiation of green-space accessibility and identify service blind zones. The XGBoost model is adopted to predict areas suitable for green-space construction, and the NSGA-III algorithm is applied to realize collaborative multi-objective optimization covering service efficiency, ecological benefits, and land development costs. The results reveal that the 15-min walking coverage of community parks in central Changsha only reaches 57.29%. Respectively, 34.52% and 41.04% of residential communities record accessibility levels below the municipal average of urban parks and forest parks, with prominent shortages of green-space supply in peripheral urban areas. This study optimizes and screens twenty-eight candidate sites for community parks, twelve candidate sites for urban parks, and eight candidate sites for forest parks. The proposed scheme effectively narrows the gap in green-space accessibility across the whole city and coordinates ecological conservation with land development costs. Compared with research relying on a single model or two-stage coupling frameworks, this paper constructs a systematic workflow spanning supply–demand status assessment to multi-objective layout decision-making, enabling differentiated optimized layout of multi-tiered parks. The integrated framework effectively enhances the spatial resilience and resource utilization efficiency of urban green-space systems, facilitates high-quality and sustainable upgrading of urban living environments, and provides a referable innovative approach for multi-level urban park arrangement and refined multi-objective planning.