Data-Driven Residence Energy Consumption Prediction Model Considering Water Use Data and Socio-Demographic Data

Data-driven energy prediction models can help urban planners and policymakers evaluate urban energy consumption patterns and then make informed decisions on how to improve urban energy efficiency. Typically, energy use data and building characteristics data were used to train these data-driven models. Few research utilized water use data and socio-demographic data, which have nexus with the energy consumption of buildings. This research utilized energy use data, building characteristics, socio-demographic, and water use data of multi-family buildings in New York City to train Least Absolute Shrinkage and Selection Operator (LASSO), Ridge Regression (RR), Support Vector Regression (SVR), and Random Forest Regression (RFR) machining learning models. The effects of socio-demographic and water use features on the performance of energy prediction were analyzed. Results showed that water use feature had significant positive impacts on the performance of LASSO, RR, and RFR models. Socio-demographic features had obvious positive impacts on the performance of SVR and RFR models. RFR trained with the BW dataset (including building characteristic features and water use feature) performed the best.