Analyse an email
Flags spam when confidence ≥ 70%
How the browser mode works
When no backend is configured, the page applies keyword-driven heuristics calibrated from the original model to estimate the probability an email is spam.
- Spam indicators: classic trigger words ("free", "urgent", "verify") yield higher scores.
- Ham indicators: business-friendly language ("meeting", "report", "thanks") reduces false positives.
- Threshold: adjustable via
APP_CONFIG.spamThreshold(default 0.70).
Deploying the Flask backend restores the full Random Forest pipeline and probability calibration logic.