Various location specific attributes cause segmentation of the housing market into submarkets. The question is, whether the most relevant partitioning criteria are directly related to the transaction price or to other, socio-economic and physical, features of the location. On the empirical side, several methods have been proposed that might be able to capture this influence. This paper examines one of these methods: neural network modelling with an application to the housing market of Helsinki, Finland. The exercise shows how it is possible to identify various dimensions of housing submarket formation by uncovering patterns in the dataset, and also shows the classification abilities of two neural network techniques: the self-organising map (SOM) and the learning vector quantisation (LVQ). In Helsinki, submarket formation clearly depends on two factors: relative location and house type. Price-level clearly has a smaller role in this respect.