Sea Ice Challenge

Kaggle & Data Visualisation challenge for Climate Crisis AI 2021 Hackathon developed by the great Vel Ice team

With Random Forests
Overall RMSE: 2.6

With Neural Network
Overall RMSE for velocity magnitude: 4.58
Overall RMSE for velocity direction: 1.28

With XGBoost
Overall RMSE for velocity magnitude: 4.27
Overall RMSE for velocity direction: 1.21

1- Installation

Install the libgeos-dev library

linux:

sudo apt-get install libgeos-dev

MAC (I have not tested this)

based on https://stackoverflow.com/questions/42299352/installing-basemap-on-mac-python 1- some home brewing

brew install matplotlib
brew install numpy
brew install geos
brew install proj

2- Download Basemap source tar file (https://sourceforge.net/projects/matplotlib/files/matplotlib-toolkits/), untar it
3- Add "export GEOS_DIR=/usr/local/Cellar/geos/3.5.0/" to a new line in .bash_profile, and then reload it via:

source ~/.bash_profile

4- From within untarred Basemap directory:

python setup.py install

Windows

follow https://matplotlib.org/basemap/users/installing.html

Python requirements

Make sure you have python 3 installed as the default version of python then in a command window or terminal run:

python -m pip install -r requirements.txt 

2- Run the software

Data import, manipulation and plotting

run sea_ice_data_plot.ipynb

Data correlation and PCA analysis

run sea_ice_data_PCA_analysis.ipynb
run sea_ice_data_visualization.ipynb

To view the plots in the "data visualization" notebook, follow this link:https://nbviewer.jupyter.org/github/Climate-Crisis-AI-Team-Vel-Ice/sea-ice-challenge/blob/master/sea_ice_data_visualization.ipynb

Training

sea_ice_data_ML_XGB_RF.ipynb