Fig. 3
From: A comparative study: classification vs. user-based collaborative filtering for clinical prediction
a A schematic of a local neighborhood for collaborative filtering of size k=3. The distance between individuals and patient 9 is quantified by a cosine distance. b The predicted recommendations for patient 9 are aggregate estimates over the neighborhood. In this simple example, the missing values are simply calculated as the average across the neighborhood, although in practice, the calculation is weighted by similarity