Credit Risk Assessment

A machine learning web application that assesses credit risk for loan applicants using an ensemble of gradient boosting models. Built with Dash and deployed on Render.

---

Live Demo

credit-risk-assessment-de3b.onrender.com

---

Project Structure

├── artifacts/
│   ├── model_1.pkl                  # Voting Classifier (GB + XGBoost + LightGBM)
│   ├── model_2.pkl                  # Random Forest (backup model)
│   ├── confusion_matrix.png
│   ├── roc_curve.png
│   ├── feature_importance.png
│   ├── feature_importance.csv
│   └── predictions.csv
├── data/
│   ├── credit_risk_dataset.csv      # Raw dataset
│   ├── train.csv                    # 80% stratified split
│   └── test.csv                     # 20% stratified split
├── notebooks/
│   ├── EDA.ipynb
│   ├── modeling.ipynb
│   └── web_application.ipynb
├── src/
│   ├── prepare_data.py              # Cleaning, outlier clipping, train/test split
│   ├── preprocess_data.py           # Imputation, encoding, scaling
│   ├── train_models.py              # Model training and artifact export
│   ├── web_app.py                   # Dash web application
│   └── assets/
│       └── styles.css
├── requirements.txt
└── README.md

---

Models

Model 1 - Voting Classifier (primary) A soft-voting ensemble of three gradient boosting models:

• Gradient Boosting Classifier
• XGBoost
• LightGBM

Model 2 - Random Forest (backup) Used alongside Model 1 to produce a consensus-based risk verdict.

Risk is determined by comparing both model outputs:

• Both predict default → HIGH RISK
• Both predict no default → LOW RISK
• Models disagree → MODERATE RISK

---

Features Used

Feature	Description
Age	Applicant's age
Annual Income	Yearly income in ZAR
Employment Length	Years in current employment
Loan Amount	Requested loan amount in ZAR
Interest Rate	Loan interest rate (%)
Loan-to-Income Ratio	Loan amount as a fraction of income
Credit History Length	Years of credit history
Loan Grade	Lender-assigned grade (A–G)
Home Ownership	RENT / OWN / MORTGAGE / OTHER
Loan Intent	Purpose of the loan
Default on File	Prior default history (Y/N)

---

Running Locally

1. Clone the repo

git clone https://github.com/yourname/credit-risk-assessment.git
cd credit-risk-assessment

2. Install dependencies

pip install -r requirements.txt

3. Prepare data and train models (skip if using pre-built artifacts)

python src/prepare_data.py
python src/train_models.py

4. Start the app

python src/web_app.py

Then open http://localhost:8050 in your browser.

---

Deployment (Render)

• Build Command: pip install -r requirements.txt
• Start Command: gunicorn src.web_app:server

---

Dependencies

dash
pandas
plotly
gunicorn
scikit-learn
joblib
xgboost
lightgbm