I am looking for an algorithm that outputs a breakdown of which features contributed the most towards a data point being labelled as an outlier.

It can be supervised or unsupervised.

At the moment, I'm using isolation forest which is really good at spotting the outliers in my data set but it's really hard to figure out why they were picked when you have hundreds of features.

I'm using python/pandas/etc.

  • 2
    $\begingroup$ Maybe look at Cooks distance for each single feature? $\endgroup$ – Peter Feb 25 '20 at 17:08

Your Answer

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

Browse other questions tagged or ask your own question.