Welcome to Insurance Predictor’s documentation!
This project demonstrates a MLOps pipeline for deploying a machine learning model into a production-ready web application. The goal is to help an insurance company forecast patient charges using input like age, gender, BMI, number of children, and smoking status.
Technical Architecture
System architecture overview
Machine Learning Model
Algorithm: Ensemble model using PyCaret AutoML framework
Training Data: US Insurance dataset with demographic and health features
Target Variable: Insurance charge amount (USD)
Model Format: Serialized pickle file for production deployment
Web Application
Backend: Python Flask application serving predictions via REST API
Frontend: Responsive HTML/CSS interface with JavaScript validation
API: JSON-based endpoints for programmatic access
Server: Production-ready Waitress WSGI server for concurrent requests
Containerization & Deployment
Container: Docker image based on Python 3.11 runtime
Registry: Azure Container Registry for secure image storage
Orchestration: Azure Web App for container hosting
CI/CD Pipeline
Version Control: Git with GitHub for repository management
Automation: GitHub Actions workflows for build/test/deploy
Testing: Pytest-based unit test suite covering API endpoints
Security: Encrypted secrets for cloud resource credentials
Contents
- Installation
- App Deployment to Azure
- API
- Unit Tests
- Documentation