Search code examples
vectorcluster-analysislocaloutlierselki

ELKI Outlier detection on 1D data

I have been trying to get LOF on a 1D data based on Eucledian distance. But I keep getting "Could not evaluate outlier results, as I could not find a minority label." Error. Please see below. The data looks like this: 0.366959 0.134065 0.54 0.292419 0.449071 0.42 0.208460 0.336666

Here is the command I give:

 -dbc.in C:\Users\...\Adata.txt -algorithm clustering.em.EM,outlier.lof.LOF -em.k 3 -lof.k 3

Clustering is done correctly but it fails on outlier detection. Am I missing something here?

Solution

  • That is not an error, but a warning.

    ELKI tries to automatically evaluate, but can only do so if it has labeled data.

    Try writing the output to a file. It probably worked, and you just didn't notice because it doesn't have a nice scatter plot visualization for 1D data.

    -resulthandler ResultWriter -out someFolder