Comparing Rough Set Theory with Multiple Regression Analysis as Automated Valuation Methodologies

Abstract

This paper focuses on the problem of applying rough set theory to mass appraisal. This methodology was first introduced by a Polish mathematician, and has been applied recently as an automated valuation methodology by the author. The method allows the appraiser to estimate a property without defining econometric modeling, although it does not give any quantitative estimation of marginal prices. In a previous paper by the author, data were organized into classes prior to the valuation process, allowing for the if-then, or right “rule” for each property class to be defined. In that work, the relationship between property and class of valued was said to be dichotomic. A real estate property may be considered inside or outside a specific class of valued. This paper introduces a valued tolerance relation to allow for more flexible rules, and offers an objective measure of discriminant threshold. In this case, the results have been derived from an explicit and specific relationship. The methodology was tested on 600 transactions in the residential property market of Helsinki, Finland. The sample of property transactions were divided into two parts (wherever possible) to calculate both internal validity on 390 “in-sample” property transactions and valuation accuracy in an “out-of-sample” group of 210 property transactions, thus obtaining interesting results.