Ali, UsmanUsmanAliShamsi, Mohammad HarisMohammad HarisShamsiBohacek, MarkMarkBohacekPurcell, KarlKarlPurcellHoare, CathalCathalHoareO'Donnell, JamesJamesO'Donnell2024-06-062024-06-062021-09-03978-1-7750520-2-92522-2708http://hdl.handle.net/10197/26167The 17th International Building Performance Simulation Association (IBPSA) Conference, Bruges, Belgium, 1-3 September 2021Urban planning and development strategies are undergoing a transformation from conventional design to more innovative approaches in order to combat climate change. As such, city planners often develop strategic sustainable energy plans to minimize overall energy consumption and CO2 emissions. Planning at such scales could be informed by spatial analysis of the building stock using Geographic Information Systems (GIS) based mapping. A data-driven methodology could aid identification of building energy performance using existing available building data. However, existing studies in literature focus on either a single building or a limited number of buildings for energy performance prediction, thus, ignoring multiple scales. This paper develops a methodology for GIS-based residential building energy performance prediction at multi-scale using a data-driven approach. The machine-learning algorithm predicts building energy ratings from local to national scale using a bottom-up approach. The multi-scale mapping process integrates the predictive modeling results with GIS. This study demonstrates the methodology for the Irish residential building stock to evaluate the energy rating at multiple scales. Modeling results indicate priority geographical areas that have the greatest potential for energy savings.enData drivenMachine learningUrban energy modelingGISEnergy performance certificateConference Publication10.26868/25222708.2021.301772022-08-0515/spp/e3125https://creativecommons.org/licenses/by/3.0/ie/