Predictive Model
A backend-wrapped Streamlit project that loads the original saved model artifacts and returns annual cost predictions with uncertainty and feature drivers.
Inputs
6
age, BMI, children, sex, region, smoker
Best model R2
0.846
XGBoost comparison metric
Backend
FastAPI
wraps original shared.py
overview
The backend imports the original Insurance Cost Predictor Streamlit shared module and calls make_prediction directly. It reuses the saved subgroup Random Forest regressors, smoker classifier path, quantile regression interval, and model comparison metrics.
role
Portfolio integration: replaced the Streamlit UI with a Next.js dashboard while preserving the original inference path and model artifacts.
backend runner
Calls /api/insurance/predict on the FastAPI wrapper at http://localhost:8000.
original predictor inputs
Model selector
Block 2 stratified Random Forest
backend contract
Start the backend with uvicorn backend.main:app --reload --port 8000. The UI displays mode and provenance so recorded artifacts are clearly distinguished from live computation.
Configure the inputs and run the backend to display the original project trace and outputs here.
architecture flow
The live pipeline trace appears in the backend runner after execution. This section shows the original project components that the backend wraps.
classifier
Uses known smoker status directly, or estimates smoker probability for unknown status.
model
Loads the original smoker and non-smoker Random Forest regressor artifacts.
uncertainty
Uses the original q10/q50/q90 quantile models for the 80% prediction interval.
explainer
Ranks local drivers using feature importance times deviation from dataset medians.
tools and models
classifier
Uses known smoker status directly, or estimates smoker probability for unknown status.
model
Loads the original smoker and non-smoker Random Forest regressor artifacts.
uncertainty
Uses the original q10/q50/q90 quantile models for the 80% prediction interval.
explainer
Ranks local drivers using feature importance times deviation from dataset medians.
example input
Age 35, BMI 27.5, one child, female, northeast region, unknown smoker status.
final result
The backend returns the original Block 2 routed prediction, q10/q50/q90 interval, subgroup costs, smoker probability, model comparison, and feature-impact table.
limitations