Energy Efficiency Analysis in Residential Buildings using Machine Learning Techniques

Vibhu Sharma

Abstract: Residential buildings account for a significant portion of global energy consumption, emphasizing the pressing need to optimize their energy efficiency. This paper proposes a novel approach utilizing advanced machine learning algorithms to accurately predict energy performance in residential buildings. Leveraging a dataset sourced from the UCI Machine Learning Repository, comprising diverse building shapes simulated in Ecotect with varying glazing areas, distributions, and orientations, our study aims to forecast critical real-valued responses pertaining to energy efficiency. Through meticulous preprocessing and feature engineering, coupled with state-of-the-art machine learning techniques such as Autogluon, PyCaret FLAML, and AutoSKLearn, our research demonstrates promising results, highlighting the transformative potential of machine learning in informing sustainable architectural practices.

Keywords: Energy efficiency analysis, Residential buildings, Machine learning techniques, Feature engineering, Predictive modeling, Sustainable architecture, Data-driven approach, Advanced algorithms, Preprocessing, Model evaluation