Other use cases

While this may be interesting, most of us don’t hunt revolutionaries as part of our day jobs. So, what good is this? Well, there are lots of uses for doing predictions against networks. Lately, graph ML has gotten a lot of interest, but most articles and books tend to showcase models built by other people (not how to build them from scratch), or use NNs. This is fine, but it’s complicated, and not always practical.

This approach that I showed is lightweight and practical. If you have network data, you could do something similar.

But what other use cases are there? For me, the ones I’m most interested in are bot detection, and the detection of artificial amplification. How would we do that? For bot detection, you may want to look at features such as the age of the account in days, the number of posts made over time (real people tend to slowly learn how to use a social network before becoming active), and so on. For artificial amplification...