Segmentations De Marches Et Decoupages Spatiaux : Quelques Applications Des Methodes Neuronales A L'Ile-De- France

– The Kohonen's self organizing algorithm has been used to divide the Paris agglomeration area into homogenous districts, designed to capture the spatial dynamics of the city. The same method was applied to three distinct sets of data, concerning the real estate market conditions, the public housing sector and the regional labor market, In each case, the method resulted in a specific spatial model of the Paris agglomeration. After a short presentation of those three segmentations, we consider the correspondances between them. It can be reasonnably argued that the variety of classification for a same parish is justified by its special caracteristics regarding the level of price, the kind of population leaving there, the type of public housing dwellings, or its position in the spatial division of labor. On the whole, a non-ambiguous hierarchy can be established among the 287 parishes we have been able to classify. Then, we explain in section 4 of the paper how those segmentations enlighten the spatial structure of the town, We first examine the logics of spatial discrimination within the agglomeration. Discrimionant analysis shows that we can speak of a parisian system of differenciation, among which the most active variables are the real estate price system, the caracteristics of dwellings and the family structure of households. A specific set of variables may be identified for differentiating the urban zones as distinct pieces of the metropolitan labor market. As regards its spatial dynamics, the parisian urban space exhibits a two-ways movement : an enlargement of the main discrepancies between the spatial components of the urban structure, and a general revaluation, notably consisting in a raising of housing real prices, an enrichment of jobs, and an accelerated distortion of the social structure towards middle-class categories. Unemployment and public housing occupation are the only caracteristics according to which we can observe an absolute devaluation of the low range components of the parisian spatial structure.