Use machine learning to improve your data quality.
Our developers can provide your company with Location Based Machine Learning. We can assist you with creating deeper learnings by combining your data with numerous open datasets.
One of our new Machine Learning products is LoBaML for housing. After noticing the extreme deviation of transaction prices and the WOZ value we started a hobby project to experiment if we could do this better. The main goal of this project was to learn to use machine learning and not create a product. After showing our results to partners/clients/people we decided on making this a product for the company. This is a real ‘product seeks clients’ approach, other then all other assignments in StraTopo.
Our data model estimates realistic transaction prices of houses, at this moment in Limburg. The model uses unique geofactors as inputs, which are generated by means of complex geodata models within StraTopo. We already have promising results and we only just started. There is a lot of room for optimization, focus and expansion. This data model has already turned into a excellent and reliable machine learning product.
Comparing transaction prices, our estimated results and the WOZ value we already achieved a lower mean absolute error compared to the WOZ value.
We use a postgres database with the postgis and pgrouting extension. With these extensions we can run the geo data models which provide unique input features.