Credit Risk Evaluator ML Model
End-to-end credit risk scoring system aligned with real-world financial standards
Domain / Function
Finance, Risk Analytics & Machine Learning
Project Overview
Built a complete Credit Risk Evaluation System that predicts and classifies loan applicants into Poor, Average, Good, and Excellent risk categories, inspired by industry-level credit scorecards such as CIBIL.
The system processes customer, loan, and bureau data, performs extensive preprocessing, feature engineering, model training, and delivers real-time predictions through an interactive Streamlit web application.
Real-time Credit Risk Prediction 1
Real-time Credit Risk Prediction 2
Real-time Credit Risk Prediction 3
Key Features
- Credit scorecard-based risk classification
- End-to-end ML pipeline from data cleaning to deployment
- Feature engineering using financial ratios
- Class imbalance handling using SMOTE-Tomek
- Hyperparameter tuning with Optuna
- Real-time prediction using Streamlit UI
Project Details
Multiple ML models including Logistic Regression, Random Forest, and XGBoost were trained and evaluated. The final model achieved excellent performance with a Peak KS score of 85.70, demonstrating strong separation between good and risky borrowers.
This project simulates real-world banking workflows and can be extended for loan approval, risk monitoring, and financial decision automation.