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
correlation of public and private score
published 26 Jan 2021, 08:24

Hello to everyone who is curious about me jumping from rank 26 to 8. I analyzed it and I want to say here that is purely accidental. It was one of my very first submits when I was just trying around.

My first “serious” submit was minimizing the score on the training set with six fixed ambulances. It got pubic score 40.421, private 44.288. After this, I tuned this with some heuristics, choosing new positions for every time-slot. That brought me down to 40.165 / 44.083.

At least for my submits, the correlation of public and private score seems quite good, except the “winning” submit with 40.616 / 43.832.

Accidental optimisation of accident locations 😂...

Congratulations 🎊

Yes, I agree because I've jumped from 83 to 7 place)))

Wow... Was ur private lb best score same as public best score?

No, my best public score is 41.057 (private 45.219), best private is 43.546 (public 41.753).

Hopefully, you kept the code .. I have a bad habit of discarding previous code and keeping only the current best score. Lesson learnt

You are right! It was very hard for me restore the code... It took one whole day to find the seed and other conditions. Now I've also learnt the lesson)))