ABSTRACT
The particular aim of the exercise documented in this chapter is to compare house prices with an affordability indicator constructed in simple terms, that is, income. In the outputs of the quasi-dynamic analysis with the self-organising map (SOM) each feature map represents the transactions in one year. These can be analysed with respect to three direct dimensions: price, land area and location of the transaction. Each feature map can also be analysed in terms of other, latent dimensions, such as topographic features of the terrain or special architectural/design characteristics of the house or neighbourhood. The SOM output ought to be supported by domain knowledge and expert information. The analysis of feature maps generated by the SOM has an intuitive appeal: non-linearity is allowed for, real relations between items are preserved despite a reduction of dimensions, and the outcome – patterns and clusters – can be examined both numerically and visually. The time-windows approach furthermore allows for some quasi-dynamics.
