An integrated data mining-based approach to identify key building and urban features of different energy usage levels

Building and urban geometries, prerequisite at the design phase, are the key determinants of building energy consumptions. However, the key building and urban features of different energy consumption levels is rarely studied. This study proposed a data mining-based method to explore the significant building features of different building groups. In this approach, clustering classifies buildings into three clusters according to energy consumption, and the clustering results contribute a base for principal component analysis (PCA) and random forest (RF) to discover key building features affecting different energy consumption levels. To demonstrate the availability of the framework, it is applied to on a city dataset in China. The results indicate that the key geometric features for low and medium residential energy consumption clusters are Orientation, HW-South, HW-West, HWNorth and HW-East, while the key determinants for high residential energy consumption cluster are Orientation and HW-South. The key features for public buildings are similar to those for residential buildings with exception of HW-East. The findings provide insights into the key influence geometric features of different building energy usage levels, which can guide the passive design of urban buildings to efficiently reduce energy consumptions at design stage.