|
DETECT provides
several means of measuring performance of both the Risk Engine and
Patterns:
The percentage of
fraud transactions detected for a given Alert Rate.
The percentage by
value of fraud detected for a given Alert Rate.
The percentage of
fraud cases detected for a given Alert Rate.
The false-positive
ratio (FPR): the ratio of false to true positives. The FPR measures
the ratio of incorrect alerts to correct alerts. From an operational
point of view we want the FPR to be as low as possible.
ROC curves and
Precision-Recall curves
Gini - the area
above the diagonal of a ROC graph.
For more information on ROC graphs please see the Techniques Wiki.
| 
