Insurance Cost Predictor is a small ML app where I tried to solve a practical question: “roughly how much might insurance cost for this profile?”
The main goal was learning by building. I wanted to understand which inputs influence price the most, and how an end-to-end regression pipeline behaves in a real app.
Tech Stack
Python · Pandas · NumPy · Scikit-learn · Gradio
How It Works
- Takes user inputs like age, BMI, smoking status, and other profile fields.
- Applies preprocessing for numeric and categorical features.
- Runs inference with a trained regression model.
- Returns a predicted insurance cost in a simple UI.
Deployment
Live demo: https://huggingface.co/spaces/SajibXD/insurance-cost-predictor
App interface preview.
Challenges & Learnings
- Feature preprocessing matters as much as model choice for stable predictions.
- Model metrics can look fine while individual predictions still need sanity checks.
- A lightweight interface (Gradio) is great for quickly sharing ML work.
Next Improvements
- Add clearer input validation and edge-case handling.
- Track prediction confidence or uncertainty bands.
- Compare multiple regression models in the UI.