The dynamic evolution of the real estate market, as well as the sophistications of the interactions of the actors involved in it have caused that, contrary to classical economic theory, the real estate market is increasingly being thought of as a set of submarkets. This is because, among other things, the modeling of a segmented housing market allows, on the one hand, to design housing policies that are better adapted to the needs of the population, but on the other hand, it allows the generation of both marketing and supply strategies Oriented to specific population sectors. Such strategies in theory should behave as options with relatively low uncertainty, thus representing an attractive offer to all market players. However, in praxis, the segmentation of the real estate market is usually modeled on the offer. It is therefore that this paper proposes a modeling from observed preferences3 seen through intraurban migrations. In particular, it is proposed to model the market through the interaction value of Coombes, scaling the results in order to visualize the resulting submarket structure from the construction of a PAM (Partitioning Algorithm Medoids).