580 California St., Suite 400
San Francisco, CA, 94104
Figure 2 – uploaded by moustafa badawy
Figure 2 Collaborative via Content Recommendation Recall that the user’s content-based profile contains weig hts for the concepts that indicate that a user will like/dislike an object. When computing Cosine similarity between two profiles, any concept in one profile but not in the other is treated as having a weight of 0 in the other profi e. As in collaborative filtering, the prediction made for an item is determined by a weighted average of all users’ predi profi ctions for that item, using the similarity between es as the weight. This is demonstrated in Figure (2).
Related Figures (3)
