A data-driven approach to identifying cost-effective retrofits, predicting energy ratings, and evaluating national retrofit CO2 savings in UK homes

• Evaluated 41 retrofit methods using 4.8 million UK home EPC records. • Identified that cheaper retrofits are efficient way to reduce CO 2 emissions per pound spent. • Used advanced machine learning models to estimate building energy ratings with 70% accuracy. • Used Monte Carlo modelling to anticipate national CO 2 reduction. • Developed a scalable framework for retrofit policy and sustainability planning. Energy efficiency is an important factor contributing to the sustainability and for reducing energy costs. There has been an increasing attention in residential energy performance, but detailed studies exploring cost-effectiveness analysis, predictive modelling, and adoption modelling are still lacking. This study addresses these issues by analysing a large Energy Performance Certificate (EPC) dataset in the UK in 2024, having over 4.8 million property records. The research explores retrofit costs and impact data to investigate three critical research questions. First, we evaluated the energy efficiency and CO 2 savings per pound spent across different property types in the UK, analysing 41 retrofit improvement types using statistical analysis. Second, machine learning models were trained to predict a building’s energy rating from its efficiency and structural traits. Third, standard retrofit interventions were assessed for defining the actual CO 2 savings by integrating retrofit adoption probabilities and Monte Carlo simulations. Our results show that the highest energy efficiency per pound spent could be achieved with inexpensive improvements like low-energy lighting, installing hot water cylinders and draught proofing. The Voting Classifier model (XGB + RF) achieved the best discrimination with 70.8% outperforming XGBoost (69.4%), Random Forest (69.09%), and MLP Neural Network (59.5%). The simulations based on different adoption scenarios demonstrate that even a small increase in the adoption rates can lead to significant national CO 2 reductions. Overall, this study provides a transferable methodology that combines cost-effectiveness analysis, predictive analysis, and retrofit adoption modelling for sustainable housing research in the UK. The findings offer insightful applicability to guide retrofit priority, policy targeting, and future studies in sustainable residential energy planning.