Can you elaborate? I know a lot of security teams will have large amounts of machine data thrown into ELK or Splunk, but that seems like qualitative data and so there's not a lot of number-crunching to do.
Graph Analysis => Saw some guys turn the Android codebase into a graph and use that to turn a dozen minor exploits into a chain that gave them root. Pwn2own IIRC.
Statistics => Great for detecting anomalies / understanding how to evaluate manipulation of cyber adjacent systems. For example, understanding the beta distribution lets you figure out how someone will game ratings in your app store to beat out legitimate apps with similar sounding ones. And of course all of these things cut both ways: if you're on red team it helps you masquerade more effectively.
Recommenders => Obviously useful to understand from attack detection, spam filter evasion, etc.
Linguistic analysis => De-anon attackers by the language they use. Figure out automatically which email accounts have been owned by sudden changes in speech usage.