This project aims to detect fraudulent transactions in e-commerce credit data using machine learning. The workflow includes data analysis, preprocessing, model building, evaluation, and explainability using SHAP.

• Fraud_Data.csv: Contains transaction and user information with a 'class' label indicating fraud.
• IpAddress_to_Country.csv: Maps IP address ranges to countries for geolocation analysis.
• creditcard.csv: Standard credit card fraud dataset with a 'Class' label.

fraud-detection-ecommerce-credit/
  ├── data/                  # Raw data files
  ├── notebooks/             # Jupyter notebooks for analysis
  ├── output/                # Output files and results
  ├── scripts/               # (Optional) Python scripts
  ├── src/                   # Source code for preprocessing, etc.
  └── README.md              # Project documentation

1. Clone the repository and navigate to the project directory.
2. Install dependencies (recommended: use a virtual environment):

   pip install pandas numpy matplotlib seaborn scikit-learn shap missingno
   # For XGBoost or LightGBM, install as needed:
   pip install xgboost lightgbm

3. Download the data files and place them in the data/ directory.
4. Open the main notebook:
   • notebooks/01_data_analysis_preprocessing.ipynb

• Launch Jupyter Notebook:

  jupyter notebook

• Open the notebook and run cells sequentially.

• Handle missing values (impute/drop)
• Data cleaning (remove duplicates, correct types)
• Exploratory Data Analysis (EDA)
• Merge datasets for geolocation
• Feature engineering (transaction frequency, time-based features)
• Data transformation (class imbalance, scaling, encoding)

• Train-test split
• Model selection: Logistic Regression (baseline), Random Forest or XGBoost (ensemble)
• Model evaluation: AUC-PR, F1-Score, Confusion Matrix
• Model comparison and justification

• Use SHAP to interpret the best model
• Generate summary and force plots
• Discuss key drivers of fraud

• Python 3.7+
• pandas, numpy, matplotlib, seaborn
• scikit-learn
• shap
• missingno
• xgboost or lightgbm (optional, for ensemble models)

• Ensure all data files are present in the data/ directory before running the notebook.
• For large datasets, ensure sufficient memory and processing power.

This project is for educational purposes.