Feeds

Predicting Behavior

When we’re asked what we do at UDL, we have a simple answer: We enable data driven decision-making in urban environments. Most of our clients have neither (geo)data in house, nor do they know how to work with such data - but all of them are experts in their business with years of experience. Many customers moreover have large datasets with addresses that can support the development of statistic models. One example of how we enable expert driven strategies and machine-learning based strategies, is the micro-rating for commercial use, i.e. a definition whether a site is a suited for commercial use or not.

The “expert based approach” is a definition of ratings (1-5) based on predefined indicators. For example, the expert might see good locations to be close to stations and the city center or close to population with high income. Our web application provides remote access to such information and allows to interactively create ratings.

Figure 1 shows such as rating for the region of Zurich as created in the app. It has been developed together with Immobilien Basel-Stadt as two-step process. The definition of weights has been supported by our data scientists, who reviewed moving patterns of enterprises and households in Switzerland with statistic approaches.

An alternative method is a machine_learning based strategy for automatic definition of such ratings. Using our machine learning framework, we generated a probability layer, showing which locations are following typical patterns of user segments. Similar to the statistic evaluation of weights, we calibrated the model on Swiss enterprise data. The result is an index defining how suited a location is for commercial use. As compared to the expert model, it uses all available data in our database, including our #urban_morphology.

Figure 2 gives a visual impression of the city of Zurich-Altstetten. It shows the high granularity of the predictions as compared to traditional expert models. The results have directly been deployed to our Web app and the Featuer_API, i.e. it is available for all of Switzerland in 25m resolution. 

We believe by building such meaningful interpretations of urban environments and providing them to professionals, we can support in making better decisions to urban environments.

Think of a use case for your business? Mail us for an appointment or for a trial!

#UrbanAnalytics #UrbanMorphology #MachineLearning #Swissstartup #proptech #insurtech


2022-09-22 15:44