Apport de la classification floue c-means spatiale en géographie : essai de taxinomie socio-résidentielle et environnementale à Lyon

Abstract

Unsupervised classification methods are common in geography, but the most widely used such as the Hierarchical Clustering Algorithm (HCA) and the k-means are not designed to work with spatial data. New developments in spatial statistics have brought new algorithms that take space into account. For example, the ClustGeo method is the spatial extension of the classical HCA. At the same time, little attention was given to spatial fuzzy classification method in geography. The paper aims to present the SFCM method, which is a spatial extension of the fuzzy c-means. To do so, we applied this method to socio-environmental data of the agglomeration of Lyon in France. We thus compared the results to its non-spatial counterpart (FCM), to the HAC and the ClustGeo method. The results showed that the SFCM combined both advantages of the fuzzy and spatial approach, making the interpretation of the results easier.

Keywords

• spatial analysis
• fuzzy logic
• automatic classification
• equity
• clustering
• k-means