ClearRate: Personal Lines Rating Engine

Project Type: Insurance Pricing | Actuarial Modeling | GLM | Dashboard
Tools Used: Python, Pandas, Streamlit, Matplotlib, NumPy
Dataset: Simulated Personal Lines Rating Factors (Driver Age, Vehicle, Territory)

Link to Live Dashboard

Objective

Build a modular, transparent personal lines rating engine capable of calculating insurance premiums through a multiplicative model. The system integrates actuarial credibility adjustments using a Poisson log-link GLM and provides a sensitivity analysis suite to visualize how specific risk variables impact the final rate.

Key Features

• Modular Rating Core:
  • Developed a RateTable module that utilizes Pandas for efficient CSV-based factor lookups across multiple dimensions (Age, Territory, Vehicle Value).
  • Implemented a purely multiplicative premium calculation logic within a dedicated rating_engine.py library.
• GLM Rate Adjuster:
  • Engineered a credibility-weighted rate adjustment module using a Poisson log-link model.
  • Allows for the refinement of base rates based on historical loss experience and frequency/severity distributions.
• Sensitivity & Impact Analysis:
  • Built a SensitivityAnalyser to programmatically test variable shifts and quantify their effect on total written premium.
  • Visualizes “one-way” and “multi-way” impact charts to ensure rating factor logic remains actuarially sound.
• Interactive Underwriting UI:
  • Streamlit-powered dashboard (app.py) that enables users to input driver profiles and instantly see a transparent premium breakdown.
  • Includes real-time visualization of loss-cost curves and factor relativities.

Deployment Highlights

• High-Performance Architecture: Avoided heavy SQL dependencies by leveraging optimized Pandas vectorized operations for sub-millisecond premium generation.
• Streamlit Cloud Deployment: Hosted the interactive engine for accessible stakeholder testing and actuarial review.
• Clean Code Principles: Structured as a package with distinct modules for data ingestion, mathematical modeling, and UI rendering.

🔗 View Blog Post: Link
🔗 View Code on GitHub: GitHub Link