True Home

This MVP was built over a few weeks by the ever talented data scientist @Guy Freeman and myself. About the machine learning model: - Valuations for over 2 million homes - Random forest machine learning model - Model needed 180GB of RAM on Google Cloud to run - Accuracy: 6% median error rate Our biggest challenge was getting quality data for homes. The Hong Kong government (our data source) does not keep detailed records of properties. For example, we don't know how many beds and baths are in each unit. However, because the housing stock in Hong Kong is more homogeneous than in most parts of the world, we were still able to generate a median error rate comparable to #Zillow and #Redfin off-market valuation models. Over the last ten years, our model shows that the value of Hong Kong homes increased by a whopping 211% - mostly due to Chinese mainland buyers and the lack of developable land. The median Hong Kong home size is only 526sqft while the median price is $714,952 USD - making Hong Kong real estate the most expensive in the world.