本研究旨在利用建築分析軟體模擬並分析 12 種不同類型的建築物，探討影響能源負荷需求的關鍵因素，並建立精確的預測模型。透過建築特徵與形狀組合，生成 768 種建築形狀，分別分析冷負荷（Cooling Load, CL）與熱負荷（Heating Load, HL），兩者分別代表維持濕熱環境穩定與室內溫度舒適所需的能源需求核心。我們收集多項建築特徵，包括玻璃面積、屋頂面積、地板面積、牆體材料、建築高度與長寬比等。這些參數在不同建築中差異顯著，並深刻影響能源需求，例如玻璃面積影響熱傳導與日照吸收，屋頂面積與材料決定保溫與散熱性能。為系統化分析其對冷、熱負荷的貢獻，本文採用特徵工程方法，進行數據清理、篩選與標準化，提取與能源負荷高度相關的關鍵特徵。
在方法層面，研究進一步應用監督式機器學習技術，採用多種迴歸模型建模與預測，包括線性迴歸（Linear Regression）、隨機森林迴歸（Random Forest Regression）、支援向量迴歸（SVR）與梯度提升機（GBM），以捕捉建築特徵與負荷需求的線性與非線性關係。為提升準確性與穩定性，進一步採用堆疊式策略（Stacking Strategy）整合多模型預測結果，並以交叉驗證檢驗泛化能力。模型評估使用均方誤差（MSE）、平均絕對誤差（MAE）與決定係數（R²）等指標，全面衡量預測精度與穩健性。結果顯示，本方法可有效預測能源負荷，並為建築設計與能源管理提供可靠依據。
在應用層面，研究成果可協助設計師於建築初期選擇合適的材料、結構與形狀，優化冷、熱負荷配置，降低能源浪費與運行成本。同時，本方法補足傳統建築設計在經濟性、材料選擇與節能策略上的不足，並可推廣至智慧建築、能源調度及低碳城市規劃，促進環境友善與永續發展的建築實踐。

This study simulates and analyzes twelve building types with 768 shape variations to identify the critical factors affecting cooling load (CL) and heating load (HL). Building parameters such as glazing area, roof and floor area, wall material, height, and aspect ratio were collected, as they strongly influence building energy performance. For instance, glazing significantly affects heat transfer and solar gain, while roof characteristics determine insulation and heat dissipation.Feature engineering techniques—including data cleaning, selection, and standardization—were applied to extract the most relevant predictors. Supervised machine learning methods (Linear Regression, Random Forest, Support Vector Regression, Gradient Boosting Machines) were adopted to capture both linear and nonlinear effects. A stacking strategy and cross-validation further enhanced accuracy and generalization. Model performance, measured by MSE, MAE, and R², confirms that the proposed framework can achieve accurate and robust predictions of building energy loads.The findings provide practical guidance for early-stage building design by optimizing material selection, structural forms, and geometries to reduce energy waste and costs. The framework also supports smart building development, energy scheduling, and sustainable urban planning, contributing to low-carbon and environmentally responsible architecture.

Chen, T., & Guestrin, C. (2016). XGBoost: A scalable tree boosting system. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794. https://doi.org/10.1145/2939672.2939785
DECODE. (2023). Data-driven energy consumption prediction leveraging historical data and environmental factors in buildings. arXiv preprint, arXiv:2309.02908.
Ferrando, M., et al. (2021). Urban building energy modeling (UBEM) tools: A state-of-the-art review of bottom-up physics-based approaches. arXiv preprint, arXiv:2103.01761.
Gao, Z., Zhang, Y., Wang, X., & Li, N. (2019). Prediction of building energy consumption using improved machine learning algorithms. Energy and Buildings, 199, 58–67. https://doi.org/10.1016/j.enbuild.2019.06.012
He, Y., et al. (2017). A deep learning framework for short-term power load forecasting. arXiv preprint, arXiv:1711.11519.
Kalogirou, S. A. (2000). Applications of artificial neural-networks for energy systems. Applied Energy, 67(1–2), 17–35. https://doi.org/10.1016/S0306-2619(00)00021-9
Kwok, Y., & Lee, J. (2011). Building cooling load prediction by support vector machine with artificial neural network model. Energy and Buildings, 43(6), 1327–1336.
Lee, J., & Cho, Y. (2021). National-scale electricity peak load forecasting: Traditional, machine learning, or hybrid model? arXiv preprint, arXiv:2107.06174.
Marino, D. L., Amarasinghe, K., & Manic, M. (2016). Building energy load forecasting using deep neural networks. arXiv preprint, arXiv:1610.09460.
Mellado, J. P., Senent, M. L., & Cristóbal, J. (2020). Feature selection for building energy consumption prediction: A case study. Energy Reports, 6, 342–350. https://doi.org/10.1016/j.egyr.2019.11.033
Mishra, S., et al. (2020a). Deep-learning-based, multi-timescale load forecasting in buildings: Opportunities and challenges from research to deployment. arXiv preprint, arXiv:2008.05458.
Mishra, S., et al. (2020b). Deep learning based load forecasting: From research to deployment – Opportunities and challenges. arXiv preprint, arXiv:2008.05458.
Mohammad, F., et al. (2018). Short term load forecasting using deep neural networks. arXiv preprint, arXiv:1811.03242.
Others. (2022). Machine learning for smart and energy-efficient buildings. arXiv preprint, arXiv:2211.14889.
Sievers, J., et al. (2021). GridLearn: Multiagent reinforcement learning for grid-aware building energy management. arXiv preprint, arXiv:2110.06396.
Sievers, J., et al. (2023). Secure short-term load forecasting for smart grids with transformer-based federated learning. arXiv preprint, arXiv:2310.17477.
Sina, A., Mosavi, A., & Ardabili, S. (2022). Systematic review of deep learning and machine learning for building energy. arXiv preprint, arXiv:2202.12269.
Wang, S., Yan, C., & Xiao, F. (2019). HVAC system optimization and energy management using machine learning techniques: A review. Energy and Buildings, 199, 263–272.
Wolpert, D. H. (1992). Stacked generalization. Neural Networks, 5(2), 241–259.
Yu, Z., & Prasanna, V. K. (2013). Energy consumption prediction with big data: A review. IEEE Transactions on Industrial Informatics, 9(3), 1584–1597.
Zhao, H., & Magoulès, F. (2012). A review on the prediction of building energy consumption. Renewable and Sustainable Energy Reviews, 16(6), 3586–3592.