kelly-behavior-predict

You can run precision, recall, and F1-score backtests against a deterministic mock dataset to evaluate rule-based prediction heuristics. This offline approach ensures data privacy and reproducibility without needing live production data.

A user-behavior funnel tracks drop-off rates across stages like browse, search, compare, booking, and completion. You can visualize these segment-level funnels locally using a Hono-based server and vanilla frontend with deterministic mock data.

Yes, you can analyze funnel behavior and test prediction rules entirely offline. The dashboard operates locally using a Hono-based server and vanilla frontend, ensuring data privacy and reproducibility without external network dependencies.

Yes, the prediction dashboard supports human-in-the-loop review of rule triggers. This allows you to audit deterministic if-else logic for next-action predictions per user segment before deploying recommendations.

Using deterministic mock data limits prediction analysis to reproducible, non-sensitive scenarios. It cannot capture live production variances, meaning the backtested precision and recall metrics reflect seeded datasets rather than real-world user behavior.