Risk Modeling · ML
Fraud Detection under Class Imbalance
Severely imbalanced fraud data made baseline classifiers unreliable.
Approach
Benchmarked classifiers with SMOTE/oversampling and precision-recall evaluation to improve minority-class detection.
Architecture
Technical highlights
- 01Classifier benchmarking under severe class imbalance
- 02SMOTE and oversampling workflows
- 03Precision/recall trade-off tracking
Stack
Pythonscikit-learnimbalanced-learnXGBoost
Related case studies