Understanding the influence of building characteristics on enhancing energy efficiency in residential buildings: A data mining based study

The building sector holds a significant share in energy consumption and its impact on various aspects of human life necessitates the identification of factors influencing the energy consumption pattern. In this study, a data mining-based approach is proposed to explore the factors which can be adapted to conserve energy in buildings. In this approach, clustering classifies similar households in a group, and then the clustering results constitute a base for decision tree and association rule mining technique to discover factors affecting energy consumption. This will reduce the complexity in interpreting the results and helps in identifying the effective factors in detail with high accuracy. To demonstrate the applicability of the framework, it is applied on the U.S. household energy dataset. The results indicate that households who do not own houses show many unfavorable behaviors with respect to energy consumption. The insulation policies are properly applied only to the newly built homes with a large floor area. Highly energy consumed households with many windows often have single-pane windows, while low-consumption energy households have double/triple glazing windows. The research findings indicate the ability of the approach to discover patterns and unknowns in more detail and precision.