Exploring influential parameters affecting residential building energy use: advancing energy efficiency through machine learning

Rapid development and escalating energy demand of India’s residential sector are crucial for sustainable development and climate mitigation. Therefore, decarbonizing the residential building sector is essential to realize India's net-zero future by 2070. Accurate energy consumption prediction and a clear understanding of influential features can yield substantial energy savings and decision-making toward efficient building design and city energy planning. Machine learning (ML) is recognized as a promising strategy for predicting energy demand and building benchmarking. Hence, this study predicts the energy consumption of residential buildings by collecting and analyzing 596 household surveys of Jaipur city in India. Six machine learning algorithms including linear regressor (LR), ridge regressor (RR), Bayesian ridge regressor (BRR), passive regressor (PR), Lasso regressor (LSR), and support vector regressor (SVR) are utilized to predict the energy consumption by considering three characteristics of dataset such as sociodemographic (SD), built structure (BS), and energy measures (EM). Furthermore, the study applies Shapley Additive exPlanation (SHAP) for the selection of influential features and analyzes their impact on energy consumption. The outcomes of the study reveal that BRR is the most efficient predictive model for SD and EM datasets, achieving an improvement in R² by 52.97% and 2.11% while reducing the MSE by 70.17% and 38.53%, respectively. Whereas SVR is most reliable for BS datasets, improving in R2 by 34.73% and 59.37% reduction in MSE. The model summary results show that the energy consumption depends on sociodemographic, built structure, and energy measures features by 77.62%, 82.55%, and 99.42%, respectively. Furthermore, income level, built-up area, and air-conditioning energy consumption are highly influential features for energy prediction. Based on the findings, this study provides energy-saving recommendations for both newly constructed and existing buildings to achieve energy savings. Moreover, this research discusses the key insights for policymakers and designers to optimize building design and implement targeted energy-saving policies to achieve energy efficiency in residential buildings.