The Risk calculator computes the safety of the roads of Geneva. It can easily be generalised to other Swiss Cities and could also be used worldwide with a bit more of an effort.

Few preliminary notes:

  • Not all the roads of Geneva are in our dataset, just the ones where there were accidents and traffic data, but begin entering a road name and let the autocompletion guide you!
  • The names that appears are lower cased and with no special characters

Search for a road... or feel lucky !

Choose a race and the results will display here.

Here are some statistic about the risk

Min Risk Max Risk Mean Risk Quartile (25%) Median (50%) Quartile (75%)
1.0 6.0 1.893 1.277 1.671 2.185

The riskiest roads of Geneva are the Route du mandement , Route de Chancy and Route de Saint Julien .

To compute the risk, we asked Google the nearest road to each of our geolocalised accident. We then matched those roads with the roads of the traffic dataset found on Swiss Data so that we could have the traffic associated to each accident. This process took several days since Google has a limit of how many "questions" it will answer per day (for free). We then agregated the accidents per road to have a notion of risk per road.
But then we realised that if there is a lot of traffic per road, it is quite expected that the number of accidents is also greater. That is why we decided to normalize the risk by the traffic. We computed this risk for several sub-categories of instances involved e.g. only for accidents involving pietons, bikes and so on.
Finally, for the risk to be more "attractive" we rescaled it from 1 to 6, which the one you have seen above (only for accidents involving cars).
We decided not to show the roads for which we had no data (either no accident or no traffic) such that it is explicit that we do not have data for thoses cases, instead of putting the risk to the minimum (one). Moreover, note that the risk have been computed only using the accidents of 2015 for technical reasons seen in the notebooks, but can be extended with almost no efforts.

For further details (in particular about the computation of the risk), please refer to the Jupyter Notebook.
Moreover, here the risk has been computed only for the incidents involving cars, but it has been computed for many other categories in the above mentioned Notebook.