Implementation of an algorithm described in this paper. It’s using LOF from scikit-learn where the two parameters being optimised are
- n_neighbors and
- contamination
The algorithm is trying to find a parameter combination that gives most clear-cut outliers.