Course Objective & Scope
Fraud data is extremely imbalanced, labels arrive late and fraudsters constantly change behaviour. The programme covers building behavioural features such as transaction velocity, device and network signals, supervised models with XGBoost/LightGBM, hybrid approaches combined with anomaly scores, and working alongside rule engines.
Evaluation uses thresholds set by review capacity and loss-amount metrics; concept drift and feedback delay are discussed. Participants prepare a scoring model and an operational decision-threshold proposal on a case adapted to their sector.
Who Should Attend?
- Fraud analytics and risk teams
- Data scientists in banking, payments, e-commerce and insurance
- Fraud operations and rule management specialists
- Prerequisite: Python and experience with classification models
Key Takeaways
- Engineer behavioural and velocity features that carry fraud signal
- Train and calibrate gradient boosting models on extremely imbalanced data
- Combine rule and model scores in a single decision flow
- Set thresholds based on review capacity and loss amount
- Build a monitoring plan that accounts for drift and delayed labels
Curriculum
Fraud Types and Data Structure
- Card, account takeover, application and insurance fraud
- Label delay and class imbalance
Behavioural Feature Engineering
- Velocity, device, location and session features
- Deviation-from-profile metrics
Supervised and Hybrid Models
- Scoring models with XGBoost/LightGBM
- Combining anomaly scores and rule engines
Evaluation and Threshold Management
- PR-AUC, recall at capacity and loss-amount metrics
- Alert queues and review prioritisation
Drift, Monitoring and Feedback
- Shifting fraudster behaviour and concept drift
- Retraining loop driven by analyst decisions
Other Courses in This Category
-
Variable analysis with WoE/IV, logistic regression scorecards, gradient boosting alternatives, reject inference and score calibration for credit, insurance and supplier risk, building explainable and auditable risk scores.
- Live Course
- 5 modules
-
Churn, purchase and cross-sell propensity models covering target definition, customer behaviour features, survival analysis, uplift modelling and campaign targeting, turning predictions into measurable retention and sales actions.
- Live Course
- 5 modules
-
Collaborative filtering, content-based and hybrid recommenders, matrix factorisation, implicit feedback, cold-start strategies and ranking metrics for building personalised product and content recommendation engines.
- Live Course
- 5 modules
-
Failure probability and remaining useful life (RUL) prediction from sensor, telemetry and maintenance logs, with hands-on signal feature engineering, health indices, early-warning thresholds and maintenance cost optimisation.
- Live Course
- 5 modules
-
SKU-location demand forecasting covering promotions, price elasticity, stockout censoring adjustment, intermittent demand (Croston/ADIDA), hierarchical reconciliation and probabilistic forecasts tied to safety stock.
- Live Course
- 5 modules
-
Bridging predictive models and mathematical optimisation: LP/MILP formulation, capacity, scheduling, routing and budget/price allocation with OR-Tools and Pyomo, plus stochastic and contextual decision systems under uncertainty.
- Live Course
- 5 modules
Get in touch
