Churn prediction, one request at a time.

About

What it does. Given contract type, tenure, charges, and service flags, the API returns churn probability, a binary prediction (F1-tuned threshold from training), and a risk band.

How it was built. Python pipeline: feature engineering, Logistic Regression / Random Forest / Gradient Boosting, best model by ROC-AUC. This deploy is inference-only — weights live in model/churn_model.joblib.

• GET /meta — stack, dataset, training summary
• GET /health — API + model file check

Hold-out test metrics

Best model by test ROC-AUC. F1 @ 0.5; live API uses F1-tuned threshold.

Key findings

From EDA and modeling — full notebooks in the repo.

• Class balance. ~27% churn; stratified split.
• Who churns. Month-to-month and shorter tenure dominate.
• Models. Three candidates compared; API serves ROC-AUC winner.
• Threshold. F1-tuned on the test set — not fixed 0.5.

Chart missing. From repo: python scripts/generate_figures.py, copy reports/figures/model_comparison.png → vercel_demo/public/model_comparison.png.

Notebooks on GitHub — 01 EDA, 02 features, 03 models, 04 threshold.

Limitations

• Dataset. Public teaching data — not production distribution.
• Churn definition. Snapshot label; real goals may need different cutoffs.
• Ops. No retraining, drift, or monitoring in this demo.