Loan Default Prediction System

End-to-end machine learning project that predicts loan default risk and serves predictions through both a web UI and a REST API.

Why This Project

Financial institutions must assess loan risk accurately. This project focuses on production-ready ML engineering: reproducible training, modular pipelines, and deployable inference services.

What It Does

• Predicts default risk (0 = unlikely, 1 = likely)
• Returns probability of default
• Exposes inference via FastAPI and Streamlit

Tech Stack

• Python, Pandas, NumPy, scikit-learn, Joblib
• FastAPI (REST API), Streamlit (interactive UI)
• Docker, GitHub Actions

Architecture (High Level)

• Data ingestion: load and split raw loan data
• Data transformation: preprocessing and feature encoding
• Model training: train and persist best classifier pipeline
• Prediction pipeline: load persisted pipeline for inference
• API + UI: FastAPI endpoints and Streamlit app

Quickstart

1) Create and activate a virtual environment

python -m venv venv
venv\Scripts\activate

2) Install dependencies

pip install -r requirements.txt

3) Train the model

python -m src.pipelines.train_pipeline

4) Run the FastAPI app

uvicorn app:app --reload --host 0.0.0.0 --port 5000

Open:

http://127.0.0.1:5000

Docs:

http://127.0.0.1:5000/docs

5) Run the Streamlit app

streamlit run app_streamlit.py

API Example

curl -X POST http://127.0.0.1:5000/predict ^
  -H "Content-Type: application/json" ^
  -d "{\"Age\":35,\"Income\":500000,\"LoanAmount\":200000,\"CreditScore\":600,\"MonthsEmployed\":60,\"NumCreditLines\":3,\"InterestRate\":12.5,\"LoanTerm\":36,\"DTIRatio\":0.9,\"Education\":\"Bachelor\",\"EmploymentType\":\"Salaried\",\"MaritalStatus\":\"Single\",\"HasMortgage\":\"No\",\"HasDependents\":\"Yes\",\"LoanPurpose\":\"Car\",\"HasCoSigner\":\"No\"}"

Docker

docker build -t loan-default-app .
docker run -p 5000:5000 loan-default-app

CI Notes

CI validates imports and application startup while skipping pipeline load if model artifacts are missing.

Project Structure

src/
  components/   # ingestion, transformation, training, prediction
  pipelines/    # training entrypoint
templates/      # HTML UI
static/         # CSS assets
app.py          # FastAPI app
app_streamlit.py

Screenshot

Author

Ellias Sithole