Comparison with S-ESD and S-H-ESD


#1

After some searching I came across Twitter’s AnomalyDetection algorithms: Seasonal ESD and Seasonal-Hybrid ESD which are extensions of Extreme Studentized Deviate aka Grubbs’ Test: https://en.wikipedia.org/wiki/Grubbs’_test_for_outliers

The paper is here:

Has anyone ran the NAB against this method? It came out in April 2017, so long after then original NAB paper.


#2

I would love to see another entry in NAB! I don’t know of anyone who has worked on this.


#3

Is this the same Twitter technique that’s listed on the NAB scoreboard now and in our more recent paper? That was based on this blog by the same author as in the arXiv paper.


#4

Thanks for pointing that out. I probably didn’t see it in your tables since you abbreviated it to ADVec.


#5

I’m glad you pointed out the arXiv paper - I had only seen the blog to date. The paper is a better reference than the blog post!