"Neural network" for identifying pests on the site?

Question

PHP

Collects, analyzes and issues a ban to the user if he believes that he is a scammer, or harms the site. What do you advise? Who has already done something similar or worked with something similar? Libraries? Sources?
Let's say a user is spamming, on a form, blog, on his wall, the neural network fires the whole thing and issues a ban. Or mothers of forum participants in their own language, etc. It is unrealistic to manually monitor everything, even if you put filters, there will be a lot of filters, and it’s unrealistic to put on everything.
Conventionally, there will be a harmfulness coefficient, let's say 100%, this is finally an inveterate pest. The network reduces or increases the coefficient. Let's say a user has spammed on a forum, and his maliciousness is 20%, the network gives a mut on the forum, blog, etc., the user cannot comment. The user starts to write everything in a personal and send spam + 50%, blocking some more actions. At 90%, you can’t do anything, just walk around the site and write to those support. If it starts to do some kind of muting, such as parsing the site, creating a load and an obstacle to its performance, everything also needs to be monitored, manually you yourself understand what functionality is, and this is just a basic example. However, after some time, the coefficient starts to drop if the user deletes his spam, and starts, for example, fumbling posts in the connected twitter account about our site. It brings profit

Answer 1

It is unrealistic to manually monitor everything, even if you put filters, there will be a lot of filters, and it’s unrealistic to put on everything.

What do you think a neural network is? Button - make beautiful?

Answer 2

Not from spam, without neurons, but it might be useful
https://habr.com/company/southbridge/blog/255343/