A novel weighted loss tab transformer integrating explainable AI for imbalanced credit risk data sets

This project presents a comprehensive Credit Risk Prediction System, leveraging an ensemble of advanced machine learning models for backend analysis while offering a seamless user experience through a web-based frontend. The system is built using Flask and MySQL, enabling users to register, log in, and predict their credit risk status. In the backend, multiple algorithms including Random Forest, AdaBoost, CatBoost, and a Stacking Classifier are trained to enhance model robustness and prediction accuracy. These models collectively analyze critical financial attributes such as credit history, loan amount, employment status, and savings. However, for simplicity and faster response time, the Random Forest model is deployed at the frontend to provide real-time credit risk predictions, categorizing users into 'Low Credit Risk' or 'High Credit Risk'. This hybrid strategy ensures high reliability from backend model training while maintaining efficiency in user interaction. The system supports effective risk management by assisting banks and financial institutions in identifying potential high-risk applicants. The modular design allows easy updates, scalability, and potential future integration of more sophisticated ensemble models. Overall, this project demonstrates the practical application of machine learning for secure, scalable, and user-friendly financial risk assessment.   Keywords: Credit Risk Prediction, Random Forest, Stacking Classifier, AdaBoost, CatBoost, Machine Learning, Flask, MySQL, Financial Risk Management, Ensemble Learning