colum2131 / mlcas2021teamqu

MLCAS2021teamQU

This is the code for MLCAS2021 Crop Yield Prediction Challenge.

The directory structure is shown below.

├─input
│ ├ Training
│ ├ Test Inputs
│ └ clusterID_genotype
├─exp006
├─Dockerfile
├─docker-compose.yml
└─.dockerignore

After installing docker and docker-compose, launch the container with the following code.

docker-compoese up --build

The code is main.ipynb in exp006.

Solution

Model

• lightGBM
• 10 seed averaging

Features

• Weather Features
  • statistical features(Maximum, Minimum, Mean, Median, Standard deviation, Skewness)
• Other Features
  • Ordinal Encoding
  • Count Encoding
  • Target Encoding(only Genotype ID, Location)
• Combine multiple features(Combine other features into a new variable)
  • Ordinal Encoding
  • Count Encoding

Cross Validation

• StratifiedKFold(Genotype ID)