Contractor Misconduct

Blogpost on this project: https://medium.com/@lukepersola/are-government-contractors-that-use-lobbyists-better-at-avoiding-penalties-215ec8e2c2f5

I worked on this project at Metis.

Contents

• data/
  • contractor_misconduct.csv - all data from The Project On Government Oversight's Federal Contractor Misconduct Database
  • clean/ - output from notebooks
    • *indicators* - with indicator dimensions
    • the rest - compact, for analysis
• notebooks/
  • wrangle/
    • misconduct_clean.ipynb - extract from CSV source and clean
    • add_indicators.ipynb - add indicators
  • analyze/
    • class_correlations.ipynb - visualize correlations between categorical features and target
    • TSNE.ipynb
  • predict/
    • predict_disposition_type.ipynb - survey of model performance
    • predict_fine.ipynb - (work in progress) included)

Usage

Run:

1. misconduct_clean.ipynb
2. add_indicators.ipynb
3. disclosures_extract_clean.ipynb
4. add_lobbyist_features_to_misconduct.ipynb
5. what you want from analyze/ or predict/