PulseCheck: Instant, Explainable Credit Risk Scoring

A Streamlit application demonstrating instant credit risk scoring with explainable AI using XGBoost and SHAP.

🎯 Project Overview

PulseCheck is a dual-view credit risk scoring system that provides:

• Customer Portal: Simple loan application with instant decisions
• Bank Officer Dashboard: Detailed risk analysis with SHAP explanations

🚀 Quick Start

Prerequisites

pip install streamlit pandas numpy duckdb joblib shap plotly

Running the Application

streamlit run app.py

The app will open at http://localhost:8501

📊 Features

Customer Portal

• Simple loan application form
• Instant eligibility decision
• Clear explanations in plain English
• Professional banking interface

Bank Officer Dashboard

• Login with admin/admin
• View all applications with filters
• SHAP-based explainability for each decision
• Portfolio analytics and model performance metrics
• Manual override capability with audit trail

🤖 Model Performance

• Algorithm: XGBoost (Gradient Boosting)
• ROC-AUC Score: 0.9501
• F1-Score: 0.83
• Training Dataset: 32,581 loan applications

📁 Project Structure

PulseCheck-Credit-Risk-Scoring/
├── app.py                                 # Main Streamlit application
├── data/                                  # DuckDB database storage
│   └── applications.duckdb                # Application database
│   └── credit_risk_dataset.csv            # Raw CSV Data
├── models/                                # Models
│   └── xgboost_model.pkl                  # XGBoost Trained Model
|   └── scaler.pkl                         # Feature scaler
├── credit-risk-prediction-models.ipynb    # Model training notebook
├── Credit_Risk_Project_Presentation.pptx  # Presentation
├── MSDS422_PulseCheck_Report.pdf          # Report for the project
├── PulseCheckEDA.ipynb                    # Data exploration notebook
├── requirements.txt                       # Requirements
└── README.md                              # This file
└── License                                # MIT License

🎓 Academic Context

This project was developed for MS DSP 422 at Northwestern University, demonstrating:

• Binary classification for loan default prediction
• Handling class imbalance (22% default rate)
• Explainable AI implementation using SHAP
• Professional UI/UX design for financial services

💡 Key Technologies

• Frontend: Streamlit
• ML Model: XGBoost
• Explainability: SHAP
• Database: DuckDB
• Visualization: Plotly

📝 License

MIT License - See LICENSE file for details

---

PulseCheck Credit Risk Scoring System
Version 2.0 | Powered by XGBoost & SHAP