Machine Learning, Architectural Styles and Property Values

Erik Johnson and I have been working on this paper for a while now and we are happy to share it with the world… Feedback is really welcome.

“This paper first introduces an algorithm that collects pictures of individual buildings from Google Street View. Second, it trains a deep convolutional neural network (CNN) to classify residential buildings into architectural styles, taking into account spatial dependencies of these styles. Third, it investigates whether architectural styles influence house prices. For resales, the architectural style is a significant determinant of transaction prices while no such effect is found for new buildings. Additionally, we are able to provide guidance on how to detect and overcome some of the limitations of machine learning methods through a large-scale comparison of predictions and expert classifications.”


Author: thies

Lecturer (Assistant Professor) for Real Estate Finance, Cambridge