🛡️ Credit Card Fraud Detection

ML pipeline on 50,000 transactions — 0.174% fraud rate · 4 models · Random Forest achieves 98.14% AUC-ROC

Total Transactions
50,000
FY 2024 dataset
Fraud Cases
87
0.174% of total
Best AUC-ROC
98.14%
Random Forest
Precision
62.5%
Fraud class
Recall
58.82%
Fraud caught
Features Used
31
V1–V28 + engineered

🤖 Model Performance

4 Models
Precision · Recall · F1 by Model Benchmark
AUC-ROC Score Higher is Better

🧮 Confusion Matrix

Random Forest
Feature Importance — Top 10 RF
True Negative
9,982
Legit → Legit ✅
False Positive
1
Legit → Fraud ⚠️
False Negative
17
Fraud → Legit 🚨
True Positive
0
Fraud → Fraud ✅

⏰ Hourly Fraud Rate

24h Pattern
Fraud Rate by Hour of Day (%) UTC