Fraud Detection Code

• This repository contains a complete pipeline to detect fraudulent transactions using e-commerce and credit card data.

Structure

• data/raw/: Original CSV datasets.
• data/processed/: Cleaned and merged fraud data.
• notebooks/: EDA, feature engineering, model training, explainability.
• src/: Python modules for each pipeline step.

How to Run

1. Install dependencies: pip install -r requirements.txt
2. Run preprocessing: python src/preprocess.py
3. Train models: python src/train.py
4. Evaluate models: python src/evaluate.py
5. Visualize SHAP explanations: python src/shap_explain.py

Models

• Logistic Regression
• XGBoost Classifier

Metrics

• F1 Score
• ROC AUC
• Average Precision
• Confusion Matrix

Notes

• Handles imbalanced classes using SMOTE.
• SHAP is used for model interpretability.

---

Developed for Week 8 of the Adey Innovations Inc. ❤️ YN