Help with ML finding RCA from dimensions

[kuzkuz]

Hello, first steps with ML, appreciate guidance on which ML method to use to get started.

We have a set of metrics marked as Good\Bad and we want to identify a probable root cause based on the various dimensions.

• See below CSV example, there are 2 entries with bad values for metric "health"
• There are multiple dimensions OS, Server, Storage
• in this case the root cause is OS = Win7, which is easy to point out

What ML method can be used to provide the correct result?

Thanks!

Name,metric_name,result,OS,Server,Storage
PC-1,health,good,Win10,Server1,S1
PC-2,health,good,Win10,Server2,S1
PC-3,health,good,Win10,Server3,S1
PC-4,health,good,Win10,Server1,S1
PC-5,health,good,Win10,Server2,S1
PC-6,health,good,Win10,Server3,S2
PC-7,health,good,Win10,Server1,S2
PC-8,health,good,Win10,Server2,S2
PC-9,health,bad,Win7,Server3,S2
PC-10,health,bad,Win7,Server1,S2

[hkeswani_splunk · Accepted solution]

Predict Categorical Assistant could be applied on this straight ahead using MLTK or you can use classifier algorithms like Random Forest Classifier directly via spl.