Data mining to estimate urban property value

Abstract

Estimate the value of real estate properties may be considered a complex process due to several factors that can influence it - from the properties physical and surroundings characteristics to the political and economic agents. In the taxation context, an accurate estimation of the property value has a special meaning, considering that taxes raised correspond to a significant portion of a municipality's revenue. This work aims to study the association between the property value and its physical and locational characteristics, in order to estimate its value for taxation purposes. That association will be studied using data mining techniques associated to geographic information systems (GIS). Initially, it will be examined how the economic value of a property is influenced by its topological relationships with the urban elements on its surrounding area (rivers, type of roads, shopping and services, among others). Will be applied data mining techniques for variable selection, identification of homogeneous regions (neighbourhoods) and to estimate the property value itself. The results will be compared with those obtained by traditionally modelling techniques (geostatistical analysis and hedonic models). The conclusion of the work will present an overview of the issue, addressing the key benefits and limitations encountered while applying the techniques. (AU)