Online Fraud Payment Detection using Machine Learning

This project implements a machine learning model to detect fraudulent transactions in online payments. The system analyzes various transaction features to classify whether a payment is fraudulent or legitimate.

Dataset

The project uses the PaySim synthetic dataset, which simulates mobile money transactions based on a sample of real transactions extracted from one month of financial logs from a mobile money service implemented in an African country. The dataset can be found on Kaggle:

PaySim Mobile Money Transactions Dataset

Dataset Features

• type: Type of transaction (CASH_OUT, PAYMENT, CASH_IN, TRANSFER, DEBIT)
• amount: The amount of the transaction
• oldbalanceOrg: Original balance before the transaction
• newbalanceOrig: New balance after the transaction
• isFraud: Target variable indicating whether the transaction is fraudulent (1) or not (0)

Implementation

The project is implemented in Python using Jupyter Notebook and includes:

1. Data preprocessing and exploration
2. Feature engineering
   • Transformation of categorical features
   • Label encoding for transaction types
3. Model training and evaluation
   • Dataset splitting into training and test sets
   • Classification model implementation

Requirements

• Python 3.x
• Jupyter Notebook
• Required libraries:
  • numpy
  • pandas
  • scikit-learn

Usage

1. Clone this repository
2. Download the dataset from Kaggle (link provided above)
3. Open payment_detection.ipynb in Jupyter Notebook
4. Run the cells sequentially to:
   • Load and preprocess the data
   • Train the model
   • Evaluate the results

Project Structure

├── README.md
├── payment_detection.ipynb    # Main Jupyter notebook containing the analysis
└── requirements.txt          # Python dependencies

License

This project is open source and available under the MIT License.

Acknowledgments

• Dataset provided by PaySim on Kaggle
• Inspired by real-world financial transaction systems