Uber Nairobi Ambulance Perambulation Challenge
$6,000 USD
Can you use ML to create an optimised ambulance deployment strategy in Nairobi?
1029 data scientists enrolled, 331 on the leaderboard
17 September 2020—24 January 2021
130 days
#1 . Approach
published 10 Feb 2021, 16:11
edited 43 minutes later

My approach to the Uber Nairobi Ambulance Perambulation Challenge ;

To begin with, it was important to observe the possible causes of road accidents and the factors that influence these incidents.

I observed unsurprisingly that the Time of the day, Day of the week, Public Holidays, and weather conditions had a lot of influence on the accident patterns. You can check this out in more details here

Then I grouped the accident locations based on this and optimized the ambulance locations for each time of the day(3Hr window), Day of the week, Public Holidays, and extreme weather conditions using a gradient descent algorithm to minimize the closest ambulance location to each accident location.

Since these groups of data are overlapping, The tricky part was determining which group takes preference. For example, the Ambulance location obtained for All days 3-6 pm may perform better than the location obtained from a more specific set of data like Sundays 3-6 pm.


  • running each grouping individually and observing how their optimized locations differ from the general location and also how they performed during the validation and test period
  • Choice of the Validation period. After observing that accident locations were influenced by time and period. I chose a validation period that mirrored the test set.

I'll be happy to answer questions or clarifications under this thread

Thanks a lot. It might inspire others to do the same. And congrats on winning!

Thanks for sharing. Great solution

Thank you very much for sharing! Congrats!