Abstract: Systems and methods for generating recommendations based on data derived from a social network are described. For example, a community in a social network service to which a plurality of member profiles belongs may be selected. A first prediction score for a first member profile and a second prediction score for a second member profile may be generated. Each of the first prediction scores may be based on a function of a first time period in which both a source member profile and the corresponding member profile belonged to the community. A selection is then made between the first member profile and the second member profile based on a comparison of the first prediction score and the second prediction score. A connection recommendation representing the selected member profile is then surfaced to an account of the source member profile.