Stilted building is a unique kind of buildings distributed across the southwest mountainous area in China, especially the Miao and Dong dwellings in Qiandongnan area. In this paper, two local traditional stilted houses, old-fashioned and new-fashioned, were selected for case study to investigate indoor environmental quality by field survey. The indoor thermal, luminous and acoustic environment and CO₂ concentrations were tested and evaluated. Results indicated that the electric heating cannot meet the comfort requirements due to the bad thermal performance of the wooden envelope, while heating by the coal stove would cause indoor air pollution. The sound resistance effect of the wooden fabric is so poor that indoor human activities can bring the noise interference to the adjacent room. In addition, the indoor natural lighting illumination was much lower than the standard requirements due to improper interior space layout design, such as few windows with small open area. It is necessary to improve indoor environment for this traditional house for sustainable development. In this study, the indoor environmental conditions in the local traditional houses was objectively evaluated, which helps to guide the residence retrofit and improve indoor environment in the future.

Urbanization in China and the proliferation of high-rise office buildings have led to increased demand for daylighting and thermal comfort. These requirements often result in reliance on active systems, including heating, cooling, and artificial lighting, which increase energy consumption. Existing studies have often focused on individual cases or room-scale models, which makes it difficult to generalize findings to the design of various high-rise office building types. Therefore, in this study, parametric prototype building models for high-rise office buildings were developed based on surveys of completed and under-construction projects. These surveys reflected actual design practices and were used to support systematic performance evaluation and typology-level optimization. Building performance was simulated using Grasshopper and Honeybee to generate large-scale datasets, and stacking ensemble learning models were used as surrogate predictors for energy use, daylighting, and thermal comfort. Multi-objective optimization was conducted using the non-dominated sorting genetic algorithm III (NSGA-III), followed by strategy formulation. The results revealed the following: (1) the proposed prototype model establishes clear parameter ranges for geometry, envelope design, and thermal performance, offering reusable models and data; (2) the stacking ensemble model outperforms individual models, improving the coefficient of determination (R²) by 0.5–16.1%, with mean squared error (MSE) reductions of 4.4–70.6%, and mean absolute error (MAE) reductions of 2.8–45.8%; (3) space length, aspect ratio, usable area ratio, window U-value, and solar heat gain coefficient (SHGC) were identified as primary performance drivers; and (4) optimized solutions reduced energy use by 3.79–11.81% and enhanced daylighting comfort by 40.16–50.32% while maintaining thermal comfort. The proposed framework provides localized, data-driven guidance for early-stage performance optimization in high-rise office building design.