Intelligent graph based visual approach for assessing and optimising energy performance in residential buildings

Optimising house energy performance requires not only understanding individual attributes but also quantifying how their interdependencies govern overall efficiency. In our study of 36,534 UK dwellings from the Department for Levelling Up, Housing & Communities, each home was modelled as a directed network of seven core attributes (nodes) connected by six weighted edges with 0.25 network density. We applied min–max normalisation, proportional rule-based edge weighting, and an adapted Dijkstra algorithm to construct seven efficiency pathways. Polar bar charts revealed that the three highest-ranked paths achieved median cumulative impact scores above 0.75, while P-value distribution analysis confirmed that five attributes heating cost, lighting cost, CO2 emissions per floor area, and hot water cost significantly (p < 0.05) drive network dynamics. In our case studies of detached, semi-detached, and terraced homes, using typology-specific percentile cut-offs stopped features from clumping together and made it easier to spot which ones matter most. The colour-coded network views turn the tangle of relationships into a clear to-do list for retrofit, so homeowners and policymakers can focus on what works. Simple network measures, how many nodes, how many links, and how dense the graph is, scale from a single house to a whole city. Put together, the method connects fine-grained feature checks with the shape of the whole network, giving a practical, evidence-based way to target actions that cut energy use and emissions. The results showed that some attributes, such as heating and hot water costs, had stronger effects on overall efficiency, while others showed moderate or small effects. These moderate effect sizes suggest that improvements in energy performance depend on the combined influence of several factors rather than a single dominant attribute.