Feature
Local ML (Naive Bayes) — A lightweight local classifier that learns from your site’s own spam vs not-spam decisions.
How it works
Tokens are normalized (URLs/emails/numbers become special tokens). Training updates a local DB table so ML becomes one signal in your score, not the entire decision.
How to access / enable
- WP Admin → AegisSpamGuard → Settings
- Enable “Local ML” (if present) and set ML weight.
- Go to Spam Log → select events → Train Spam / Train Not Spam.
Recommended setting
Recommended: ON with a modest weight (e.g., 8–15 points).
Keeping ML as a small signal prevents model drift from overpowering strong deterministic signals like honeypot or deny rules.
Why you need this
Admins get site-specific learning without sending data to a cloud service.
Additional information