Éditeur(s) : HAL CCSD
Résumé : National audience
We propose a method to solve the problem of clustering individuals described by mixed variables structured in blocks. It is a hierarchical method with two levels similar to Hierarchical Principal Component Analysis presented by Wold, based on the mixed topology maps (MTM). The first step consists in establishing, for each block of mixed variables, a clustering representing a synthesis of local individuals at the block level. The second step is to re-apply MTM on the weighted results of the first level. We obtain a global clustering of individuals, consensus among the partitions from level 1. The proposed method is illustrated on data from the national survey in the French housing stock carried out in 2003-2005 by the Observatory for Indoor Air Quality (OQAI).
Fouille de Données Complexes, complexité liée aux données multiples et massives (FDC)
bordeaux, France

hal-00748841
https://hal.archives-ouvertes.fr/hal-00748841
https://hal.archives-ouvertes.fr/hal-00748841/document