Overview §
Built a machine learning model to classify credit card transactions as fraudulent or legitimate using Python and Scikit-Learn.
Tested multiple models (Logistic Regression, Decision Trees, Random Forests) and achieved the best results with Random Forest.
Techniques Used §
- Data preprocessing & feature scaling
- Handling class imbalance with SMOTE
- Model evaluation using ROC, Precision, Recall, F1-score
- Visualization of fraud vs. non-fraud distribution
Results §
- Best-performing model achieved:
- High Recall (> 90%) for fraud detection.
- Good Precision to reduce unnecessary alarms.
- ROC-AUC score above 0.95.