A machine learning-based predictive analytics system developed to classify fraudulent credit card transactions using supervised learning techniques, feature engineering, and anomaly pattern recognition. The model was trained on transactional data to identify fraud indicators, generate risk predictions, and improve detection accuracy for suspicious financial activities.
Click to view full document
✔ Predictive Fraud Classification
✔ Transaction Pattern Analysis
✔ Feature Engineering Pipeline
✔ Fraud Risk Scoring
| Model | Accuracy | Precision | Recall | F1 Score | AUPRC (Positive Fraud) |
ROC AUC |
|---|---|---|---|---|---|---|
| Logistic Regression | 97.9% | 98.8% | 96.6% | 97.7% | 0.992 | 0.989 |
| XGBoost | 97.4% | 98.8% | 95.5% | 97.1% | 0.998 | 0.998 |
| Random Forest | 97.4% | 100.0% | 94.3% | 97.1% | 0.969 | 0.997 |
| Deep Learning | 97.4% | 98.8% | 95.5% | 97.1% | 0.994 | 0.994 |