Automated Valuation Models : Improving Model Performance by Choosing the Optimal Spatial Training Level

The use of Automated Valuation Models (AVMs) in the context of traditional real estate valuations and their performance has been discussed in the academic community for several decades. Most studies focus on finding which method is best suited for estimating property values. One aspect that has not yet been studied scientifically is the appropriate choice of the spatial training level. The published research on AVMs usually deals with a manually defined region and fails to test the methods used on different spatial levels. The aim of our research is thus to investigate the impact of training AVM algorithms at different spatial levels in terms of valuation accuracy. We use a dataset with about 1.2 million residential properties from Germany and test four different methods, namely Ordinary Least Square, Generalized Additive Models, eXtreme Gradient Boosting and Deep Neural Network. Our results show that the right choice of spatial training level can have a major impact on the model performance, and that this impact varies across the different methods