You might be interested in the field of Process Mining -

https://en.m.wikipedia.org/wiki/Process_mining

Author here.

That's actually a really interesting field. However, it seems to rely on the ability to link events (having a trace of what happened) whereas the use-case in this article is to group individual logs based on similarities (and depending on the algorithm used, on structure).

Still, that does sound like a field I'd like to explore, thanks for making me discover it!

You may also want to check out https://www.celonis.com/ - they've been around in this field for a long time.

Wow, I just realized Celonis acquired https://www.integromat.com/ , which I followed back some time for process automation :wow:

You might want to link to https://docs.datadoghq.com/logs/explorer/patterns/ might in the "Why?" section.

Indeed! I updated the post.

Thanks.

You might be interested in a similar algorithm I've written a few months ago on patterns clustering: https://github.com/goldblat/klapa