That's a really clever way to incorporate TOR as an input, without just disabling all TOR-enabled accounts. Were you able to get a bayesian algorithm to learn

That's a really clever way to incorporate TOR as an input, without just disabling all TOR-enabled accounts. Were you able to get a bayesian algorithm to learn whether specific users were likely to be spammers or trolls? (As opposed to specific posts.) I'm guessing this isn't easy to do, since I get spammed on other services on a regular basis (e.g. Twitter spammers will setup new accounts and spam on a regular basis).