Machine Learning-Based Method for Detached Energy-Saving Residential Form Generation

In recent years, machine learning has gradually been applied to building energy-saving designs reduce the time consumption of optimization screening stage. However, since most existing research scholars come from fields computers and engineering, application technology mostly involves complex programming as well software in field which requires multiple be coupled achieve. view differences between disciplines high threshold, these theories are difficult apply promote practical work architecture. this regard, paper focuses on improvement methods, based Grasshopper platform, proposes a detached residential form generation design method process, explore optimal more concise efficient way. Based new method, basis verifying its feasibility through case, two algorithms, neural network (ANN) support vector (SVM), compared studied, applicability algorithms different performance indicators is further discussed. The results show that ANN model highest accuracy suitable for prediction energy consumption; simple fast operation SVM, it comfort with relatively low requirements. By combining above efficiency can improved while satisfying relevant indicators. This help architects quickly search best scheme stage provide data information feedback conception deepening.