WiDS Datathon 2020 Second place solution
The code, provided here, allows to generate a final submission, that scored 0.91378 on the public leaderboard and 0.91485 on the private.
- Python 3.7
- Install Kinoa from https://github.com/oleg-panichev/kinoa or just comment all related code
Detailed description is posted in Kaggle discussions.
Clone this repository:
git clone https://github.com/oleg-panichev/WiDS-Datathon-2020-Second-place-solution.git
Put the data in the 'input' directory and run the code:
cd WiDS-Datathon-2020-Second-place-solution/src
nohup python -u ensemble.py > ensemble.log &
You may find the code for each model within ensemble along with experiments log in the corresponding directories within _kinoa_ directory.
- Iryna Ivanenko ([email protected])
- Oleg Panichev ([email protected])