Seonglae ChoAlgorithm that detects abnormally high activity spikes during specific periods compared to normal levels
- Create a baseline of "average activity levels" using historical data
- Compare with current activity levels
- If significantly higher than baseline → high risk score
Example: Vendor A typically sends 100-200 invoices per quarter, but suddenly sends 500 at year-end → flagged as "anomalous"
- Card payment analysis
- Payroll analysis
- Vendor analysis