Thanks!

To answer your question: I would recommend to use Cook's distance if you have a linear regression problem (only numerical features), DBSCAN if you only have numerical (scaled) features (not necessarily linear regression), and Isolation Forest if your data also contains categorical features.