PremiumPredictor

A full-stack machine learning application for insurance premium prediction using a RandomForestClassifier model. Features REST API endpoints with health monitoring, Docker containerization, and production-ready deployment architecture using Supervisor for multi-service management.

Problem Statement

Insurance companies need accurate premium predictions to price policies competitively while maintaining profitability. Manual underwriting is slow and inconsistent, creating a need for automated ML-driven predictions.

Architecture Overview

A RandomForestClassifier model trained on historical insurance data provides premium predictions. FastAPI serves the model through REST endpoints with health monitoring. Streamlit provides an interactive frontend for users to input parameters and view predictions. Docker containerization with Supervisor enables multi-service deployment on AWS.

Technical Challenges Solved

• Feature engineering for insurance data to maximize prediction accuracy
• Building a production-ready ML serving architecture with health monitoring
• Orchestrating multiple services (API, frontend, model server) with Supervisor
• Containerizing the entire stack for consistent deployment across environments

Key Learnings

• RandomForestClassifier provides a good balance of accuracy and interpretability for tabular insurance data
• Health monitoring endpoints are essential for production ML systems
• Supervisor simplifies multi-process management in containerized environments
• Streamlit enables rapid prototyping of ML interfaces for stakeholder feedback