Using DBSCAN to find data that are far from high density clusters?

Question

Conscious that dbscan clusters don't necessarily have cluster centres, but for an anomaly detection task, I want to spot data that are outliers/away from the normal clusters. Is there a way to do this using sklearn's dbscan?

Answer 1

For anomaly detection use an anomaly detection algorithm instead.

There are siblings derived from DBSCAN for exactly this purpose. For example, LOF and LoOP and kNN and ...