Train, Evaluate, Repeat: Building a Credit Card Fraud Detection System || Leela Senthil Nathan

This talk covers three major ML problems Stripe faced (and solved!) in building its credit card fraud detection system: choosing labels for fraud that work across all merchants, addressing class imbalance (legitimate charges greatly outnumber fraudulent ones), and performing counterfactual evaluation (to measure performance and obtain training data when the ML system is changing outcomes itself). EVENT: PyData NYC 2018 SPEAKER: Leela Senthil Nathan PERMISSIONS: PyData Organizer provided Coding Tech with the permission to republish this video. CREDITS: PyData YouTube channel: https://www.youtube.com/channel/UCOjD18EJYcsBog4IozkF_7w https://www.youtube.com/watch?v=GPWB72fSY_o