https://www.nishika.com/competitions/4/summary
- Create 1180 features only from financial information.
- Make 3 predictions by LightGBM and CatBoost with 5 cross-validation.
- Netflix blending.
For more details, see approach.md.
I only use {2014-2017}/documents.csv
.
input
└── data
|── 2014
| └── documents.csv
|── 2015
| └── documents.csv
|── 2016
| └── documents.csv
└── 2017
└── documents.csv
docker-compose up --build -d
docker exec -it nishika-yuho bash
sh run.sh