Using principal components analysis and clustering technics to define typical buildings

The present paper presents a method to characterizethe typical building from a group by applyingprincipal components analysis (PCA).The method has been developed on a sampleof secondary education school buildings inGreece. The purpose is to define the typicalbuilding in order to propose generalized improvementsfor energy efficiency of the buildingstock concerned. Therefore seven variables fromquestionnaires have been analyzed: heated surface,age of the building, insulation of the building,number of classrooms, number of students,schools operating hours and age of the heatingsystem. Considering the frequency distributionof the examined variables, the typical buildingis defined as the closest to the sample median.As a first step, a principal components analysishas been applied for transforming the originalinterrelated variables into the same numberof new, uncorrelated variables called the principalcomponents in order to consider the load ofthe variables and to reduce the dimension of themultivariate problem. Then, in the principalcomponents coordinate system, the typicalschool has been identified as the closest to thesample median using Euclidean distance. Furthermore,k-means clustering technique may beapplied to classify buildings significantly for amore extensive analysis of the stock.The principal components coordinate systemhas eased the analysis of the multivariate sampleand its classification and could as well provide asignificant two-dimensional graphic of the sample.