Core workflows

This path turns events into analysis-ready data and reproducible experiments. Use it for point-in-time feature construction, fraud stress testing, model baselines, calibration, campaigns, and benchmark comparisons.

Tutorial	Time	Extras	Output
From Events to a Trustworthy ML Dataset	15–20 min	base	PIT dataset, quality diagnostics, and delayed-label tables
Investigate and Stress-Test Fraud Scenarios	20–30 min	base; plotting optional	replay, graph, scenario, camouflage, and counterfactual comparisons
Build a Reproducible Fraud Benchmark	20–30 min	base	benchmark metrics, model comparison, and reproducibility manifest

Tutorials

• From Events to a Trustworthy ML Dataset

• Investigate and Stress-Test a Fraud Detection System

• Build a Reproducible Fraud Benchmark

Next path: Production ML.