Supervised Fraud Model
Model v1.3-insurance-agent

Precision

0.0%

Recall

0.0%

Accuracy

0.0%

High Risk Predictions

0

Model Trend by Month

Confusion Matrix

Predicted →

True Positive

2

Correctly caught

False Positive

1

False alarm

False Negative

8

Missed fraud

True Negative

9

Correctly cleared

Actual ↓

Fraud Probability Distribution

Red dashed line = threshold (60%)

Top Predictive Features

Frequency across predictions

Prediction Threshold Monitor

Active Threshold

60%

Average Predicted Probability

28.1%

Predicted Fraud Cases

6

Last retrain: Not retrained yet

Top Supervised Predictions

PredictionSales IDPolicyProbabilityPredictedActualTop Features
SP-0023AG1003PL90002379.9%FraudFraudstatus transition, rules, payer identity
SP-0045AG1005PL90004573.8%FraudUnknownloan, status transition, rules
SP-0027AG1007PL90002760.7%FraudFraudrules, payer identity, network
SP-0001AG1001PL90000160.4%FraudNon-Fraudpayer identity, status transition, rules
SP-0012AG1002PL90001260.1%FraudUnknownstatus transition, rules, payer identity
SP-0034AG1004PL90003460.1%FraudUnknownstatus transition, rules, payer identity
SP-0041AG1001PL90004156.5%Non-FraudNon-Fraudstatus transition, rules, loan
SP-0013AG1003PL90001347.4%Non-FraudFraudloan, payment behavior, network
SP-0037AG1007PL90003747.2%Non-FraudFraudloan, payment behavior, payer identity
SP-0014AG1004PL90001445.6%Non-FraudUnknownrules, payer identity, network
SP-0007AG1007PL90000737.1%Non-FraudFraudpayment behavior, network, loan
SP-0043AG1003PL90004337.1%Non-FraudFraudpayment behavior, network, loan