Real Estate Investment Appraisal of Buildings using SOM

Abstract

The most commonly used methods to analyze and value real estates are econometric models. However, these models have some weaknesses that make it difficult to obtain good analyses or reliable models. First, there are the linearity assumptions, and second the problem of correlating variables. Furthermore, changes in real estate and dwelling prices in the 1990s have made it especially difficult to use econometric models. The aim of the research presented in this chapter is to show how the self-organizing map is more suitable for appraisal of the prices of dwellings and/or for wider use in real estate valuation. The reason is that the self-organizing map is a neural network technique that allows to examine non-linearity as well as providing a capability to preserve the topology and distribution of the data, which is very important in real estate valuation.¹ The application in this chapter involves the Finnish dwelling market.