Card Fraud Detection Using XGBoost

Project Overview

This project focuses on detecting fraudulent credit card transactions using machine learning. It was a final year internship project completed by me and my team during our internship at IIDT as AI-ML-DS Interns. We utilized the XGBoost algorithm for classification and prediction of fraudulent transactions.

Data Source

• Source: Kaggle.com
• Data Link: Credit Card Fraud Detection Dataset 2023

Required Libraries

• Python (3.x): Primary programming language for implementation.
• NumPy: For numerical computing and handling arrays.
• Pandas: For data manipulation and analysis.
• Matplotlib & Seaborn: For data visualization.
• Imbalanced-learn: For handling imbalanced datasets.
• Scikit-learn: Comprehensive machine learning library for various algorithms.
• TensorFlow or PyTorch: Deep learning frameworks for building neural networks (if needed).
• Keras: High-level neural networks API.
• Scikit-plot: For plotting common machine learning metrics.
• XGBoost or LightGBM: Gradient boosting libraries for effective classification.

Project Features

• Model Training: The XGBoost algorithm was used to train the model on the credit card fraud dataset.
• Feature Selection: The features used for training include anonymized transaction data labeled V1 to V28, along with the transaction amount.
• Model Deployment: A Streamlit application was developed for easy interaction with the model, allowing users to input transaction details and get predictions on whether the transaction is legitimate or fraudulent.

Implementation Details

• Data Loading: The dataset is loaded using Pandas.
• Model Loading: The pre-trained XGBoost model is loaded from a pickle file.
• Streamlit App: A user-friendly interface is created using Streamlit, where users can input a transaction ID to check the legitimacy of the transaction.
• Prediction: The model takes the transaction details, processes the data, and predicts if the transaction is legitimate or fraudulent.
• Result Display: The result is displayed on the Streamlit app, indicating whether the transaction is legitimate or fraudulent.

User Interaction

• Transaction ID Input: Users can input a transaction ID to check its status.
• Submit Button: On clicking the submit button, the app fetches the corresponding transaction details from the dataset.
• Prediction Result: The app displays whether the transaction is legitimate or fraudulent based on the model's prediction.

Skills Demonstrated

• Machine Learning: Application of the XGBoost algorithm for fraud detection.
• Data Handling: Efficient manipulation and processing of large datasets using Pandas and NumPy.
• Visualization: Use of Matplotlib and Seaborn for data exploration and visualization.
• Model Deployment: Creating an interactive web application using Streamlit.
• Team Collaboration: Working collaboratively on a real-world problem during an internship.

Conclusion

This project highlights the application of machine learning in detecting fraudulent credit card transactions. The use of XGBoost for classification, combined with a user-friendly Streamlit interface, demonstrates a comprehensive approach to solving a critical financial problem. The project showcases my proficiency in machine learning, data processing, and web application development, making it a valuable addition to my portfolio.