Ramp kit storm forecast

Authors: Sophie Giffard-Roisin, Alexandre Boucaud, Mo Yang, Balazs Kegl, Claire Monteleoni (AppStat-CDS)

The goal is to predict the hurricane evolution (24h forecast) using collected data from all past hurricanes (since 1979). New version.

Set up

1. clone this repository

git clone https://github.com/ramp-kits/storm_forecast.git
cd storm_forecast

2. install the dependancies

• with conda

conda install -y -c conda conda-env     # First install conda-env
conda env create                        # Use environment.yml to create the 'storm_forecast' env
source activate storm_forecast       # Activates the virtual env

• without conda (best to use a virtual environment)

python -m pip install -r requirements.txt

3. download the data

python download_data.py        # quick-test data for testing ~200Mb

4. get started with the storm_forecast_starting_kit.ipynb

New submissions

1. create a new submission <new_sub> by building on the existing ones

cp -r submissions/starting_kit submissions/<new_sub>

2. modify the *.py files in submissions/<new_sub> with your favorite editor

3. test the submission with

ramp_test_submission --quick-test --submission <new_sub>

4. if the job complete, you can submit the code in the sandbox of ramp.studio

License

BSD license : see LICENSE file

Credits

This package was created with Cookiecutter and the ramp-kits/cookiecutter-ramp-kit project template issued by the Paris-Saclay Center for Data Science.