Determining the location of gardens using Data Science

Our approach

The dataset that was available contained the address and coordinates of all residences.

We used data from the Key Register of Addresses and Buildings (in Dutch: BAG) through an API to verify whether a residential property was an apartment or a single-family home. We then supplemented the dataset with the cadastral map per parcel through the Land Registry API. Based on that, we calculated the centroid (the center point) of the plot.

If a plot only contained the coordinates of one residential property, we calculated the angle between the coordinates of the house and the centroid of the plot. Based on this, we determined the sun position (for example: East).

The result

We enriched the housing database with an extra column, called 'garden location'. It is now known per address whether the garden faces East, Southwest, etc.

For the data that we filtered from the dataset, we created an image archive. In the image you can see:

• the calculated wind direction as text
• a pink dot for the coordinates of the house
• a green dot in front of the centroid of the plot

The script we wrote is fully automated and requires little processing power. If new homes are added in the future, we can determine the garden location of all homes at the touch of a button.

97% reliability

To ensure high reliability of the data, we optimised the model through several automatic iterations. As a result, the script has a reliability of 97%. We determined this on the basis of a manual sample of 500 samples. The script can be applied to all kinds of coordinates (for example also to WGS 84, Rijksdriehoek coordinates, etc.).