Towards a data-driven profiles of household energy consumption in collective housing: Machine learning clustering with insights from Bab Ezzouar (Algiers)

Residential energy consumption in North African collective housing remains insufficiently characterised, despite the need for energy-efficiency policies that account for both household practices and building-related constraints. This study addresses this gap in public collective housing in Bab Ezzouar, Algiers, Algeria, using a valid sample of 380 households. Five years of electricity and natural gas billing records were converted into average annual on-site energy consumption and combined with survey-based socio-demographic, behavioural and perceptual variables, dwelling geometry, equipment ownership and envelope thermal properties. The analysis combines univariate analysis, principal component analysis and the comparative evaluation of standard clustering algorithms. K-means provided the most interpretable and internally consistent three-cluster solution, with a reasonable structure according to the Silhouette coefficient (0.513) and Davies–Bouldin index (0.71). The identified socio-technical profiles span approximately 14,200–19,100 kWh/year of annual on-site energy consumption. Cluster 03 corresponds to an affluent energy-intensive profile, Cluster 02 to a resource-constrained low-consumption profile, and Cluster 01 to a transition-oriented profile combining moderate consumption with favourable behavioural readiness. Cluster separation is mainly driven by dwelling size, equipment ownership and envelope thermal performance, particularly U-values, rather than by socio-demographic characteristics alone. A central finding is the dissociation between declared willingness to save energy and realised consumption, showing that positive attitudes do not necessarily result in lower energy use when physical, equipment-related and comfort-related constraints dominate. The findings support differentiated policy pathways: deep retrofits and demand-response measures for high-consuming households, energy-poverty support for constrained low-consumption households, and voluntary efficiency and renewable-energy programmes for transition-oriented households.