Linear and Tree-Based Intelligent Investigation of Cross-Domain Housing Features to Enhance Energy Efficiency

Abstract

Energy efficiency is a critical concern in built environment. Identifying key features that drive energy consumption is essential for optimizing building performance. Traditionally, studies have focused on single-domain datasets. These approaches overlook the potential insights gained from integrating data across different domains. This research addresses this gap using a cross-domain dataset that includes building characteristics, energy usage, and environmental factors. Feature selection techniques, including filter methods (correlation, mutual information), wrapper methods (RFE), embedded methods (Lasso, Random Forest, and gradient boosting), and dimensionality reduction are used to identify the most significant features contributing to the energy efficiency of residential properties. These techniques identify the most significant features influencing energy consumption. The findings show that cross-domain features like energy consumption, CO₂ emissions, and heating cost play a key role in predicting energy performance. By integrating data from multiple domains, the feature selection process reveals areas for energy optimization that are previously overlooked in single-domain studies. The results provide valuable insights for energy consultants, building managers, and policymakers aiming to enhance energy efficiency in residential buildings. This research highlights the importance of cross-domain data integration and offers a robust framework for feature selection. Ultimately, it contributes to more effectiveenergy-saving strategies and sustainable building practices.