The energy performance of buildings represents a major issue of the 21st century. Many solutions have been discussed to improve buildings' energy performance, but the actions to take differ from one building to another. In other words, current solutions are built on a case-by-case basis and cannot be extrapolated easily. Indeed, it is difficult to find generic
solutions due to their complexity and heterogeneity.
By placing buildings in groups and subgroups, one can define relevant energy optimization recommendations without auditing each building individually. Because initial labels are not always defined, clustering is relevant in our case. Since we seek for intrinsic similarities between groups and subgroups, hierarchical clustering is needed. Buildings are described with mixed
data. They include numerical data such as surface or number of floors, and categorical data like types of heating or insulation materials. Few clustering algorithms exist for mixed data, and even fewer are hierarchical. In this article, we present a method for the hierarchical clustering of mixed data based on pretopology.