Using Neural Networks to Estimate Constant Quality House Price Indices

This paper extends work in the field of Constant Quality Price Indexing by using Neural Networks to estimate the model. A series of housing data sets is used to develop constant quality price indices using traditional econometric techniques and using neural networks incorporating genetic algorithms. The analysis indicates that neural networks are a real alternative to the econometric methods. Introduction In a previous study the Local Government areas of Port Pirie, Stirling and Unley were used for some basic Constant Quality Indexing using an adjusted time series approach. In this paper the same locations are chosen for study. Cross sectional data are used involving all probable market transactions from July 1980 to June 1998. This data is analysed using a standard hedonic method utilising dummy variables to indicate time periods and also using a Neural Network model. The neural network model uses the same data, in the same format and uses Genetic Algorithms to optimize the model structure. The aim of this research is to determine is neural networks are a useful inclusion to the tools used for property price indexing.