Fraud Detection Dashboard
50,000 transactions from 495 accounts across 43 US cities — Hybrid Risk Scoring (ML + Rules)
Total Transactions
50,000
$14.9M total volume
Accounts
495
~101 txn per account
High + Critical
3,052
6.1% of all transactions
Rule Recall
94.7%
rules catch ML anomalies
Consensus Anomalies
190
flagged by 2+ models
Rule-Based Engine Analysis
7 expert fraud rules with severity weighting — Hybrid Score = 0.5 x ML + 0.5 x Rules
Rules Defined
7
Critical / High / Medium
Txn with 1+ Rule
61.9%
30,950 transactions
Recall on ML
94.7%
rules catch ML anomalies
Critical Alerts
24
0.05% of transactions
Top Trigger
Device Sharing triggers มากที่สุด (45.92%) — อุปกรณ์เครื่องเดียวถูกใช้หลายบัญชี เป็นสัญญาณ fraud ที่ชัดเจน
Multi-Rule
ธุรกรรมที่ trigger 3+ rules มีโอกาสเป็น anomaly สูงกว่าปกติ 12 เท่า ใช้ในการจัดลำดับ alert investigation
| Rule | Condition | Trigger Rate | Weight | Severity |
|---|
| R1 — Login Attempts | LoginAttempts ≥ 3 | 3.85% | 3.0 | Critical |
| R2 — Amount Spike | Amount Z-Score > 2 | 3.39% | 2.5 | High |
| R3 — High Ratio | Amount / Balance > 0.8 | 5.62% | 2.0 | High |
| R4 — Device Sharing | Device serves ≥ 5 accounts | 45.92% | 2.5 | Critical |
| R5 — Rapid Txn | Time since last < 1 hour | 2.41% | 1.5 | Medium |
| R6 — Multi-Location | Account active in 8+ cities | 18.79% | 2.0 | High |
| R7 — High Velocity | 3+ transactions per day | 0.12% | 1.5 | Medium |
Risk Explorer
Interactive exploration of risk factors and anomaly patterns
Risk Level:
Top Risk Accounts
Accounts ranked by average hybrid risk score — highest risk first
| # | Account | Amount | Channel | Login | Rules | Hybrid Score | Level |
|---|