They asked me to create a referral system that recommends a list of other users for any given user, using the network of all users. The network consists of nodes, and two users are friends if their nodes are connected by an edge.
Don't recommend current friends.
Base recommendations on mutual friends: A and B are both connected to the same node, but A and B are not directly connected.
Second, given a list of recommendations, return a score for each recommendation.
The score should be bounded, for example in [0, 1].
NDCG—I didn't think of it at the time...
AI was allowed: scoping plus AI implementation.
Discussion
Loading comments…