A small experiment on gradient boosting methods for wheat yield predictions (CGIAR challenge). This is a very one-shot experimentation project.
The project tree shoul look like
|-data/
|-models/
|-notebooks/
|-submissions/
The data folder must have all the data in it and the zip files already unzipped in it.
If you face some paths errors, you shoul modify the config.py
file to match your own configuration (especially the paths)
First run the eda.ipynb
in the notebooks
folder to generate the train & test .csv files (train_sampled.csv
and test_sampled.csv
).
Then run tne train.ipynb
in the notebooks
folder that generates the submission file into the submissions directory.
If you want to run it on Google colab, you have a colab_train.ipynb
for that. You only need to upload all the data to google drive (you do not need to unzip the image arrays).