🚀 Python GLM ML Pipeline

HowTo Data Science: Logistic Regression model churn probability

→ Watch it directly on YouTube

✅ Logistic Regression churn prediction
✅ FastAPI REST endpoint
✅ OpenAI summarisation for human-readable explanations

1. Install requirements:

pip install -r requirements.txt

2. Run locally:

uvicorn app:app --reload

• app:app loads the FastAPI app from app.py.
• --reload enables auto-reload for development (useful for code changes, not for production).

3. Test with:

curl -X POST http://localhost:8000/predict -H "Content-Type: application/json" -d '{
 "age": 45.0,
 "tenure": 24.0,
 "monthly_charges": 79.85,
 "total_charges": 1800.0,
 "contract_type": "Month-to-month",
 "payment_method": "Electronic check"
}'

Optionally enable external AI summaries:

export OPENAI_API_KEY="your_actual_openai_api_key"

Python_GML_ML_Pipeline/
├── app.py # FastAPI application
├── requirements.txt # Python dependencies
├── logistic_model.pkl # Trained ML model (placeholder)
├── scaler.pkl # Feature scaler (placeholder)
├── Dockerfile # Container configuration
├── .gitignore # Git ignore rules
└── README.md # This file

1. Build the Docker image:

docker build -t python-gml-ml-pipeline .

2. Run the Docker container:

docker run -p 8000:8000 -e OPENAI_API_KEY="your_actual_openai_api_key" python-gml-ml-pipeline

3. Access the FastAPI app:

• API: http://localhost:8000
• Health Check: http://localhost:8000/health
• Interactive Docs: http://localhost:8000/docs
• OpenAPI Schema: http://localhost:8000/openapi.json

Execution Flow:

1. Load trained model and scaler (using joblib).
2. API endpoint receives JSON data (new user or input data).
3. Dataframe creation & scaling for consistency with training.
4. Model predicts churn probability (or other target).
5. Returns JSON response with prediction for integration into apps or dashboards.

✅ Run locally:

uvicorn app:app --reload

✅ Run in Docker:

docker build -t python-gml-ml-pipeline .
docker run -p 8000:8000 -e OPENAI_API_KEY="your_actual_openai_api_key" python-gml-ml-pipeline

Key reasons to use FastAPI:

1. Modern async Python framework
2. Automatic OpenAPI schema & Swagger docs
3. Production-grade performance
4. Can be integrated into microservices/SaaS

Pierre-Henry Soria

Made with ❤️ by Pierre-Henry Soria . A super passionate & enthusiastic Problem-Solver / Senior Software Engineer. Also a true cheese 🧀, ristretto ☕️, and dark chocolate lover! 😋

@phenrysay pH-7 YouTube Video BlueSky

• logistic_model.pkl and scaler.pkl are placeholders. Train and export your own models using joblib.dump.
• This is an educational API prototype. The bundled three-feature artifacts were saved with scikit-learn 1.1.3 and do not match the six-field request or the pinned runtime. They are deliberately rejected; /health and /predict return HTTP 503 until compatible artifacts are supplied.
• Export trusted artifacts with the pinned scikit-learn version. The scaler/preprocessor must record the six request fields in their declared order and handle the categorical values; the classifier must use classes [0, 1]. Training data, evaluation and a matching preprocessing pipeline are not included.
• Generated summaries do not establish why a model made a prediction. With no OPENAI_API_KEY, no explanation request is sent; provider errors return a generic message.
• Run python -m unittest discover -s tests for offline API tests. Success-path doubles verify routing, not predictive quality. Production deployment still needs dependency review, real model validation and access controls.

"AI models become valuable when they’re deployable, explainable, and integrated into real products that create business value."