Machine Learning in Financial Credit Risk Assessment
summary
Describes building a production credit‑risk pipeline that combines applicant fields and credit‑agency history at scale (over £2bn lent to ~240,000 borrowers). Claims start with ~3,000 features and reduce to a few dozen via per‑feature trees, random‑forest importance and recursive feature elimination, then ensemble models (XGBoost and Keras neural nets) and model stacking boost discrimination; reported lift is a few hundredths of ROC AUC. Implements preprocessing and deployment in Python with Pandas NumPy Scikit‑Learn XGBoost Keras and an in‑house predictor pipeline that runs live scoring.
AI-written summary · about
chapters (7)
- 0:00ML in credit risk
- 3:29What is credit risk
- 7:01Data gathering
- 15:10Pre-processing features
- 18:20Building ML models
- 24:57Deploying models
- 36:17Q&A
related talks
27mData Science and ML with an emphasis on Banking...
40mCreating a Contemporary Lending Risk Management System Using Python
37mAutomated Machine Learning: Building a Pipeline
22mUsing data to innovate credit decisioning for Small-to-Medium Enterprises
01h 12mBuild Simpler Production ML Systems using Feature/Training/Inference Pipelines
48mProcessing Terabytes of Data in Credit Scoring
04h 03mTutorial - Intro to Data Science
58mNew-type Scoring Models: Analyzing Customer Actions