A geographic knowledge discovery approach to property valuation

This thesis involves an investigation of how knowledge discovery can be applied in the area Geographic Information Science. In particular, its application in the area of property valuation in order to reveal how different spatial entities and their interactions affect the price of the properties is explored. This approach is entirely data driven and does not require previous knowledge of the area applied. To demonstrate this process, a prototype system has been designed and implemented. It employs association rule mining and associative classification algorithms to uncover any existing inter-relationships and perform the valuation. Various algorithms that perform the above tasks have been proposed in the literature. The algorithm developed in this work is based on the Apriori algorithm. It has been however, extended with an implementation of a ‘Best Rule’ classification scheme based on the Classification Based on Associations (CBA) algorithm. For the modelling of geographic relationships a graph-theoretic approach has been employed. Graphs have been widely used as modelling tools within the geography domain, primarily for the investigation of network-type systems. In the current context, the graph reflects topological and metric relationships between the spatial entities depicting general spatial arrangements. An efficient graph search algorithm has been developed, based on the Djikstra shortest path algorithm that enables the investigation of relationships between spatial entities beyond first degree connectivity. A case study with data from three central London boroughs has been performed to validate the methodology and algorithms, and demonstrate its effectiveness for computer aided property valuation. In addition, through the case study, the influence of location in the value of properties in those boroughs has been examined. The results are encouraging as they demonstrate the effectiveness of the proposed methodology and algorithms, provided that the data is appropriately pre processed and is of high quality.

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

UCL (University College London)

Year dc:date.issued

2009

Author dc:creator

• Christopoulou, K.

• Department of Civil, Environmental and Geomatic Engineering

Language dc:language

en, eng