UNICEF Arm 2030 Vision #1: Flood Prediction in Malawi
$10,000 USD
Predict flood extent caused by storms in southern Malawi
1168 data scientists enrolled, 298 on the leaderboard
2 December 2019—1 June 2020
Ends in 2 months
Target variable
published 14 Dec 2019, 15:01
edited 3 minutes later

I have problem to understand target variable, I think Target variable should be rectangle where the flood has happened, not percentage of rectangle that was flooded, because if we consider so, the target variable didn't reflect neithier where nor when the flood happend.

What do you think ?

edited 1 minute later


If the target has a value greater than zero, it is safe to assume that it was flooded, and if it has zero, it is safe to assume it wasn't flooded.

So, for a start, you can build a model to first check if the square was flooded, then pass that classification itself as a feature into your model to then get the percentage of the square that was flooded. In theory, it should correlate heavily with your target.

Hi AIchemi

I understand, I m talking about the use of that kind of information, didn't reflect neithier where nor when the flood happend.

flooded, not flooded, okay, but when will be flooded I guess useful ?