Fig. 2
From: A drug recommender system for the treatment of hypertension
(1) Similar patients are filtered from the training patient database; (2) Observed treatments and outcomes are used to compute personalized outcome predictions; (3) Treatment options are ranked according to outcome predictions; The top-\(\mathrm{N}\) ranked treatments are presented to the user; (4) Using evidence-based criteria, treatment options can optionally be post-filtered (Sect. Knowledge-based labeling and post-filtering)