Attribute selection in hedonic pricing modeling applied to the Portuguese urban housing market

Abstract

One of the challenges associated with studying the housing market is related to the need to handle a high amount of variables. In this context, data mining techniques, and more specifically, feature selection methods allow the selection of relevant variables efficiently. Results from the application of eight different methodologies for feature selection with a real dataset on the urban housing market of Aveiro and Ílhavo municipalities show that we can build hedonic models with an acceptable explanatory power of housing prices while considerable reducing their complexity.