Towards an Intelligent Property Valuation Model: An Exploratory and Conceptual Study

Automated Valuation Models (AVMs) have been used by mortgage industry participants for a number of years with a limited degree of success. A lack of comprehensive data and thus restricted reliability has prevented AVMs from being fully adopted by governments, lenders, brokers, and the like. This paper presents a framework for building an intelligent property valuation model that overcomes these problems. An overview of AVM research, a survey of current AVM products and several interviews with industry participants were conducted. The findings revealed a number of important elements for optimal AVMs. This paper also provides a description of a unique dataset combining both appraisal data and public record data. This information combined with well established data mining techniques including clustering, geocoding, and ensembles provides the basis for further research towards building a new property valuation method that better meets industry needs.