What are some techniques that I can use for anomaly detection given a non-Normal distribution? I have less than twenty available observations.

  • $\begingroup$ Checkout QQ plots also with what JahKnows said.. $\endgroup$
    – Aditya
    Mar 22 '18 at 13:07
  • $\begingroup$ @JahKnows - if the offer still stands, I would like to ask for an easy introduction. $\endgroup$ Mar 11 '19 at 11:43
  • $\begingroup$ @user7677771, probably best to ask a separate question to avoid reviving old posts. But, sure! $\endgroup$
    – JahKnows
    Mar 12 '19 at 0:41

I would suggest a nearest neighbors approach. This technique is non-parametric, such that it does not assume your features follow any given distribution. The degree from which a novel instance can be classified as anomalous can set through some p-value estimation. These techniques are computationally expensive however due to your small dataset this may be well suited.

Check out:

Learning Minimum Volume Sets http://www.stat.rice.edu/~cscott/pubs/minvol06jmlr.pdf

Anomaly Detection with Score functions based on Nearest Neighbor Graphs https://arxiv.org/abs/0910.5461

New statistic in P-value estimation for anomaly detection http://ieeexplore.ieee.org/document/6319713/

You can also use more rudimentary anomaly detection techniques such as a generalized likelihood ratio test. But, this is kind of old-school.

  • $\begingroup$ I can elaborate on how these techniques work if you have difficulty with the paper. They're relatively easy concepts clouded in a lot of theory in the papers. $\endgroup$
    – JahKnows
    Mar 22 '18 at 10:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.