database

The following methods are used to collect cloud-based BIM data. In this method, architectural designs, structural properties, material specifications and energy usage of educational buildings will be generated using cloud-based BIM systems. After data collection, accuracy and consistency must be ensured by cleaning and standardizing the collected data which is called preprocessing. Moving to the next step is dimensionality reduction using “Principal Component Analysis (PCA)”, to overcome the problem of focusing on the important factors that have a significant impact on energy usage and sustainability. After dimensionality reduction comes regression analysis which is performed by Light Gradient Boosting Machine - Neural Network - Model Predictive Control (LightGBM-NN based MPC) in this case for analysis. For classification, a hybrid technique Support Vector Machine - Neural Network - Genetic Algorithm (SVM-NN-GA) is used.

本研究采用下述方法采集基于云的建筑信息模型（Building Information Modeling，BIM）数据。该方法依托基于云的BIM系统生成教育建筑的建筑设计方案、结构特性、材料规格与能耗数据。数据采集完成后，需通过清洗与标准化处理（该过程称为预处理）保障采集数据的准确性与一致性。下一步采用主成分分析（Principal Component Analysis，PCA）开展降维操作，以解决高维数据下易被冗余维度干扰、无法聚焦核心影响因素的问题，精准锁定对能耗与可持续性具有显著影响的关键因子。降维完成后开展回归分析，本研究采用轻量梯度提升机-神经网络-模型预测控制（Light Gradient Boosting Machine - Neural Network - Model Predictive Control，LightGBM-NN 基MPC）方法进行分析。分类任务则采用支持向量机-神经网络-遗传算法（Support Vector Machine - Neural Network - Genetic Algorithm，SVM-NN-GA）的混合技术。