Edit: DanielBMarkham's comment does a better job of conveying why this would be a bad approach.
Take a look at how many of the accounts there solely spam affiliate links from Amazon, are from a Twitter account, and repeatedly post the same items.
So alright, we aren't on the same page, but you're saying exactly what I just said.
Just because I used 2 different signals to prove to you that the method doesn't generate a lot of false positives, doesn't mean that looking at amazon links is a bad method.
If I was at Pinterest, I would slap a captcha on all Amazon links with affiliates until I had time for a fancier solution, and it would probably get rid of 99% of the spam.