We (Riadh & Gabriel) worked on a machine learning clustering model to group the areas that had a minimum number of 10 accidents within a radius of 150m. The data is the 2019 version provided for the French government.

We have here two use cases. The geobalisation and the geofencing.

You provide your departure and arrival addresses. We will geocode them into GPS coordinates, trace the route and intersect with all the dangerous areas. On the image, in red are the clusters that intersect the user's path. In blue are all other clusters provided by the model. This can be seen from the density of clusters in the Paris department. Normally by clicking on each cluster, you have the number of accidents aggregated in this area.

In a more improved version, it would provide a new path to the user. We could also provide this information in real time on a more suitable platform such as a phone. 

The project is fully scalable to several cities. We have reduced the study area here simply for the sake of efficiency. We could replicate this for devices like vehicles in our African cities.

You can test our application. Provide formalized addresses in Paris, otherwise crash crash grrrrr!