Challenge: Every day, thousands of property objects are being advertised. Object evaluations often require extensive analysis and physical visits. Evaluation reports can be very expensive, can take a long time to be compiled and might be vulnerable to uncertainty, e.g. when relying on unexperienced employees.
Solution: Geospin’s self-learning algorithms can reliably predict the rental price of residential and commercial properties of various sizes and features. Our algorithms can scale as desired and are even applicable to areas where there is no real-estate market data available.
The basis for model training consists of thousands of commercial property locations in Hamburg.
(Data provided by IS24)
The trained model recognizes complex spatial structures which affect the rental price — even for areas without comparative market data.
Comparing the prediction with the true prices, we can see that the error is marginal. Furthermore, precise location evaluations are possible in seconds.
„The approach highlighted how automated location evaluations can support our internal processes. A particular challenge comes from the availability of large and constantly updated datasets.“
Dr. Peter Scibbe, Union Investment Real Estate GmbH
Dr. Christoph Gebele