false positives

I got around to updating the Viktorfeed whitelist, and whilst I was at it, doing a clean-up of the database to remove all the known false positives by deleting all the movements that came from an airline that was on the list. Interestingly, this gave me an opportunity to calculate the false positive rate - as it turned out, about 7% of the entries in the DB were false positives, in other words, an accuracy over about a year of 93%. Which I don't think is at all bad for a system based on adding airlines to the whitelist as they appeared in the feed, but then again, this only covers the first phase of filtering and doesn't include a lot of once-off movements.