Credit Risk Scoring API

A simple FastAPI service that predicts Probability of Default (PD), converts it into a credit score (300–850 scale), and assigns a risk level.

1. Install Pipenv (if not installed)

pip install pipenv

2. Create Virtual Environment

From inside the project root (credit_scoring_backend/):

pipenv install fastapi uvicorn pandas scikit-learn joblib numpy

This will:

• Create a virtual environment
• Generate Pipfile
• Generate Pipfile.lock

3. Activate Virtual Environment

pipenv shell

You should now see the virtual environment active in your terminal.

2. Train the Model

Ensure the dataset exists at:

data/UCI_Credit_Card.csv

Run the training script:

python  train.py

This will:

• Train a Logistic Regression pipeline
• Evaluate performance (ROC-AUC, confusion matrix, classification report)
• Save the trained model to:

models/credit_model.pkl

3. Run the API

uvicorn main:app --reload

API base URL:

http://127.0.0.1:8000

Interactive documentation:

http://127.0.0.1:8000/docs

4. Endpoint

POST /credit-score

Example Request

{
  "LIMIT_BAL": 200000,
  "AGE": 35,
  "avg_pay_delay": 0.2,
  "credit_utilization": 0.45,
  "payment_ratio": 0.6
}

Example Response

{
  "PD": 0.18,
  "Credit_Score": 751,
  "Risk_Level": "Low"
}

Credit Score Logic

score = 850 - (PD * 550)

Risk Levels:

• High: score < 650
• Medium: 650–749
• Low: 750+

Project Structure

.
├── app.py
├── train_credit_model.py
├── data/
│   └── UCI_Credit_Card.csv
└── models/
    └── credit_model.pkl

Workflow

1. Train the model
2. Start the API
3. Send POST request to /credit-score
4. Receive PD, Credit Score, and Risk Level