Spatiotemporal analysis of building energy performance and urban morphology interactions across Chinese climate zones: A machine learning approach to predicting cooling demand in high-density districts

China's rapid urbanization has created unprecedented challenges in building energy management, with cooling demand in urban districts increasing by 340% between 2000 and 2023. This study develops a novel machine learning framework to predict district-level cooling energy consumption by integrating urban morphology parameters, microclimate data, and building envelope characteristics across five major Chinese climate zones. Using a comprehensive dataset of 2,847 residential and commercial districts from 43 cities—including Beijing, Shanghai, Guangzhou, Chengdu, and Harbin—the research employs ensemble methods combining Random Forest, Gradient Boosting, and Neural Network algorithms to establish predictive models with R 2 values exceeding 0.94. Field measurements were conducted during summer 2023 (June-August) using mobile meteorological stations, thermal imaging surveys, and energy consumption monitoring across representative districts in each climate zone. Results reveal that urban canyon geometry (H/W ratio) exerts the strongest influence on cooling demand in Hot Summer-Warm Winter zones (β = 0.68, p < 0.001), while building surface albedo dominates in Hot Summer-Cold Winter regions (β = 0.72, p < 0.001). Sky View Factor (SVF) demonstrates inverse correlation with cooling loads in high-density urban cores (r = -0.81), with districts exhibiting SVF < 0.3 consuming 47% more cooling energy than those with SVF > 0.6. The analysis identifies critical threshold values: districts with building density above 0.42 and average building height exceeding 85m experience nonlinear escalation in cooling demand. Machine learning models successfully predicted seasonal cooling energy consumption with mean absolute percentage error (MAPE) of 8.3%, outperforming traditional degree-day methods by 34%. Climate-specific regression analyses demonstrate that Guangzhou districts require 156 kWh/m 2 annually for cooling compared to 89 kWh/m 2 in Beijing and 198 kWh/m 2 in Chongqing, reflecting the combined impact of climatic conditions and urban form. Feature importance analysis reveals that façade-to-ground ratio, material thermal mass, and prevailing wind corridor alignment collectively account for 63% of cooling demand variance. The study proposes district-level intervention strategies including strategic greening (potential 12-18% cooling reduction), façade retrofitting with high-albedo materials (8-15% reduction), and urban canyon optimization (15-22% reduction). Implementation scenarios suggest that comprehensive district-scale interventions could reduce China's urban cooling energy consumption by 180-240 TWh annually by 2030, equivalent to avoiding 120-160 million tons of CO 2 emissions. This research provides Chinese urban planners and policymakers with quantitative decision-support tools for climate-responsive district design, establishes baseline performance metrics for existing urban morphologies, and demonstrates the efficacy of data-driven approaches in addressing building energy challenges at the city scale.