Skip to Main content Skip to Navigation
Conference papers

Segmentation d'un parc virtuel de bâtiments par clustering pour la rénovation énergétique

Abstract : In this study, clustering, an unsupervised machine learning technique, is used to segment a building stock into homogeneous groups in terms of descriptive attributes and energy performance. A virtual stock of residential buildings has been generated to test a developed clustering method. It consists of various geometries (shape and size) that identify different types of. The attributes of the buildings used for clustering are separated into two parts: the decision space of descriptive attributes, and the objective space of energy performance. The developed clustering method seeks to satisfy the homogeneity criterion in the two spaces. The interaction between two spaces has been achieved by dimensionality reduction using linear discriminant analysis (LDA) and self-supervision using the adjusted Rand index (ARI). The analysis of the results makes it possible to validate the approach of the method. The obtained clusters of buildings are more or less distant from the traditional typologies (for example in terms of morphology).
Complete list of metadata

Cited literature [13 references]  Display  Hide  Download
Contributor : Yunseok Lee Connect in order to contact the contributor
Submitted on : Friday, May 18, 2018 - 10:06:39 AM
Last modification on : Saturday, January 15, 2022 - 3:48:56 AM
Long-term archiving on: : Wednesday, September 26, 2018 - 2:03:26 AM


Files produced by the author(s)


  • HAL Id : hal-01795043, version 1



Yunseok Lee, Pierre Boisson, Mathieu Rivallain, Olivier Baverel. Segmentation d'un parc virtuel de bâtiments par clustering pour la rénovation énergétique. Conférence IBPSA France, May 2018, Bordeaux, France. ⟨hal-01795043⟩



Record views


Files downloads