XAirQ

Explainable machine learning for air quality

Structure of this repository

• The root directory contains prepare.sh, LICENSE.md, .gitignore etc.
• setup contains requirement files and hpc modules
• source contains all python scripts, ordered by topic (preprocessing, models...)
• test contains tests for the python scripts
• resources contain descriptive .csv files needed for analysis
• jupyter contains jupyter notebooks for full workflows
• doc contains documentation materials

Prerequisites

• Python Version >=3.8 with virtualenv package
• Debian packages: PROJ and GEOS, graphviz. Install on Ubuntu 20.04 with sudo apt-get install libproj-dev proj-data proj-bin; sudo apt-get install libgeos-dev sudo apt install graphviz. If you use a conda distribution, these packages should already be installed.

Getting started on own machine

• Clone the repository to JUWELS or your own linux machine
• Run source prepare.sh to activate the python environment
• Run ./run.sh to start scripts. Currently, only the option "test" is available here. in source/models/, there are some models you can run, e.g. with python random_forest.py

Code style

We try to write in the pep8 style!

• Take a look at the pep8 convention
• Start the analysis with ./run.sh and choose the pep8 option
• Enter your name, so all the scripts you wrote are checked
• Be sure to state an __author__ in the scripts you wish to check.

Getting started on JUWELS

• Clone the repository to JUWELS
• Run source prepare.sh to create the python HPC environment for JUWELS
• Go to /ozone-interpolation/hpc_scripts and sbatch train_xxx_.sh for submitting jobs (currently only Random Forest and Neural Network)

Authors

• Scarlet Stadtler
• Clara Betancourt

License

This project is licensend under MIT License.