Credit Fraud Analytics

Personal project analyzing credit card transaction data for fraud detection using anomaly detection techniques.
Focus: Handling highly imbalanced datasets (fraud is rare ~0.17%), evaluating unsupervised methods like Isolation Forest.

Why this project?

• Builds on my background in anomaly/outlier detection.
• Demonstrates practical data analysis skills: preprocessing, imbalance handling, model evaluation (PR-AUC, confusion matrix, etc.).
• Relevant to risk/fraud analytics in finance (e.g., banking roles).

Dataset

• Primary: Credit Card Fraud Detection (Kaggle/ULB)
  • 284,807 transactions over 2 days (2013 European cardholders)
  • 492 frauds (~0.172%)
  • Features: 28 PCA-transformed variables (V1–V28) + Time + Amount + Class (0=legit, 1=fraud)
  • Anonymized for privacy.

Alternative datasets (if you want variety):

• Credit Card Transactions Fraud Detection (2020 simulated)
• Credit Card Fraud 2023

Current Notebook

• credit_fraud_anomaly_detection.ipynb → Isolation Forest baseline + evaluation.

More to come: Compare with Local Outlier Factor, Autoencoders, or supervised models (e.g., XGBoost with SMOTE).

Setup (Colab/Local)

pip install -r requirements.txt
jupyter notebook

Or open directly in Colab:

Skills Demonstrated

· Python · Pandas · Scikit-learn · Anomaly Detection · Imbalanced Data · Visualization (Matplotlib/Seaborn)

Questions/collabs? Feel free to open an issue or reach out!

Main profile: github.com/S33mi | Casual X: @Seemi_Rauf