ML4science_project

Classification of ordinal outcomes for the analysis of injury severity using machine learning methods.

All this code was run in Python 3.8.11 with the following libraries: Numpy 1.19.2 Pandas 1.3.4 Sklearn 1.0.1 Pytorch 1.7.1 imbalanced-learn 0.8.1

Get started

In case some of the datasets (see below) are missing, generate the preprocessed versions with

python preprocess.py

Run

python run.py

to train an initial model with the parameters from params.py

File overview:

• dataset
  • 1.08 Crash Data (detail) DD.csv # overview over features
  • age_binned_preprocessed.csv # preprocessed with age as categorical binned columns
  • age_binned.csv # original dataset with age as categorical binned columns
  • area.png
  • crashdata.csv # original, uncleaned dataset
  • preprocessed_data.csv # preprocessed dataset: mean imputed and standardized (output of preprocess.py)
  • tempe_cleaneddata.csv # cleaned dataset
• cross_validation.py # functions for k-fold cross-validation
• data_exploration.ipynb # initial data analysis and overview
• evaluation.py # functions for model evaluation metrics
• model.py # models as subclasses of pytorch.nn.Module
• params.py # parameters for models + training
• preprocess.py # preprocessing pipeline: mean imputation, standardization, age binning
• README.md
• run.py # run training of model