Multi-stage filtering for recommended user connections on online social networks

One or more computer-readable non-transitory storage media embodying software that is operable when executed to:

• access a social graph comprising a plurality of nodes and a plurality of edges connecting the nodes, each of the edges between two of the nodes representing a single degree of separation between them, the nodes comprising:
• a first node corresponding to a first user associated with an online social network;
• and
• a plurality of second nodes corresponding to a plurality of second users of the online social network, respectively;
• identify, for the first user, a first set of candidate-users from the plurality of second users, wherein each candidate-user corresponds to a second user within a threshold degree of separation of the first user within the social graph;
• filter, using a first-stage filtering model, the first set of candidate-users to generate a second set of candidate-users, wherein the first-stage filtering model filters the candidate-users based on one or more edge-features;
• filter, using a second-stage filtering model, the second set of candidate-users to generate a third-set of candidate-users, wherein the second-stage filtering model filters the candidate-users based on the one or more edge-features and one or more user-features;
• rank, using a user-engagement model, the third set of candidate-users based on a feature superset, wherein the user-engagement model ranks the candidate-users based on a calculated probability of the first user connecting with the respective candidate-user;
• and
• send, to a client system of the first user, instructions for generating a suggested-friends interface for display to the first user, wherein the suggested-friends interface comprises one or more candidate-users from the third set of candidate-users having a ranking greater than a threshold rank.

A system comprising:

• one or more processors;
• and a non-transitory memory coupled to the processors comprising instructions executable by the processors, the processors operable when executing the instructions to:
• access a social graph comprising a plurality of nodes and a plurality of edges connecting the nodes, each of the edges between two of the nodes representing a single degree of separation between them, the nodes comprising:
• a first node corresponding to a first user associated with an online social network;
• and
• a plurality of second nodes corresponding to a plurality of second users of the online social network, respectively;
• identify, for the first user, a first set of candidate-users from the plurality of second users, wherein each candidate-user corresponds to a second user within a threshold degree of separation of the first user within the social graph;
• filter, using a first-stage filtering model, the first set of candidate-users to generate a second set of candidate-users, wherein the first-stage filtering model filters the candidate-users based on one or more edge-features;
• filter, using a second-stage filtering model, the second set of candidate-users to generate a third-set of candidate-users, wherein the second-stage filtering model filters the candidate-users based on the one or more edge-features and one or more user-features;
• rank, using a user-engagement model, the third set of candidate-users based on a feature superset, wherein the user-engagement model ranks the candidate-users based on a calculated probability of the first user connecting with the respective candidate-user;
• and
• send, to a client system of the first user, instructions for generating a suggested-friends interface for display to the first user, wherein the suggested-friends interface comprises one or more candidate-users from the third set of candidate-users having a ranking greater than a threshold rank.