Overview
The project aimed to develop an advanced credit risk management system for a national bank with over 40 million customers. This system adheres to the recommendations of the Basel Committee, focusing on identifying and managing credit risks across different client segments (commercial, corporate, and personal banking). The project addressed the bank’s need for an integrated software platform to monitor risk exposures, improve credit portfolio management, and comply with international financial standards like IFRS9.
- Data and Features
- The system analyzed data from three main customer segments: - Commercial Banking: Utilized 430 features derived from financial and operational data.
- Corporate Banking: Used 510 features, including detailed financial statements and market exposure.
- Personal Banking: Analyzed 360 features from individual credit histories and behavioral patterns.
- Data was collected, preprocessed, and validated to ensure accuracy and reliability.
Machine Learning Models
- Implemented three machine learning models:
- XGBoost (XGB): For handling high-dimensional data and providing robust predictions.
- Random Forest (RF): To explore variable importance and ensure generalization.
- Logistic Regression (Logit): For interpreting credit risk patterns.
- Models were trained, evaluated, and refined over a one-year period using extensive historical data
- Achievements
- Delivered a system capable of real-time risk assessment for loan approvals and portfolio monitoring.
- Achieved an impressive 93% recall and 98% F1-score, ensuring high reliability in identifying high-risk clients while minimizing false positives.
- Improved credit decision-making processes and compliance with international risk management standards.
- Impact
- Enhanced the bank’s ability to allocate capital efficiently by assessing PD (Probability of Default), LGD (Loss Given Default), and EAD (Exposure at Default) for each client segment.
- Supported optimal credit marketing strategies and improved customer targeting.
- Strengthened the bank's reputation and credibility by introducing an internal credit rating system.
Future Scope
- Integration with external economic and market data to enhance the accuracy of credit risk forecasts under stress scenarios.
- Expansion of the system to incorporate additional risk categories such as operational and liquidity risks.
This project demonstrates the value of leveraging advanced machine learning techniques and comprehensive data analysis to address critical challenges in financial risk management.