The relationship between building and urban mobility-based energy consumption and urban form: A machine learning approach

Urban energy consumption studies predominantly focus on building energy use. However, urban form has the potential to influence energy consumption not only through building morphology but also through mobility demands shaped by spatial location within the city. In different locational contexts such as central, peripheral, coastal, and transitional zones, urban form varies in terms of density, building layout, and accessibility, which affects both building-related and trip-related energy demand. Despite this, studies that jointly evaluate the combined effects of urban morphological characteristics and locational mobility dynamics on energy consumption remain limited. Addressing this gap, this study assumes that urban energy consumption should be examined not only at the building scale but also in an integrated manner with the locational characteristics of urban form. Within this framework, urban forms in Istanbul were classified into seven categories using the k-means clustering algorithm. Subsequently, a dataset consisting of morphological variables derived from 32,112 buildings was modeled to predict building energy consumption. Among ten machine learning techniques, Random Forest was identified as the most successful algorithm, achieving an R² value of 0.87. The effects of urban morphological variables were interpreted using SHapley Additive Explanations. In the next step, energy consumption resulting from trips between Traffic Analysis Zones across different transportation modes was calculated. When building-related and trip-related energy consumption were evaluated together, Open low rise was identified as the least efficient form with an average energy consumption of 186 kWh/year·m², while Open skyscraper was identified as the most efficient form with an average of 85 kWh/year·m². However, when areas dominated by building energy and those dominated by trip-related energy were examined, it was observed that the energy performance of urban forms varied depending on spatial context. The findings provide insights to inform energy-oriented urban planning and urban design approaches.