For this approach to work effectively, users’ friend lists need to correspond to tightly-knit communities in the network graph. To verify the extent to which users in friend lists form closely connected communities, we examine the normalized conductance [31] of the existing friend lists, whose value ranges from -1 to 1, with strongly positive values indicating significant community structure. Prior studies of social network graphs have found that normalized conductance values greater than 0.2 correspond to strong communities, that could be detected fairly accurately by community detection algorithms [31]. We analyzed the conductance values
for our 233 friend lists and we found a significant positive bias. Over 48% of the friends lists have values larger than 0.2, suggesting that a large fraction of friend lists could be automatically inferred from the social network.
For this approach to work effectively, users’ friend lists need to correspond to tightly-knit communities in the network graph. To verify the extent to which users in friend lists form closely connected communities, we examine the normalized conductance [31] of the existing friend lists, whose value ranges from -1 to 1, with strongly positive values indicating significant community structure. Prior studies of social network graphs have found that normalized conductance values greater than 0.2 correspond to strong communities, that could be detected fairly accurately by community detection algorithms [31]. We analyzed the conductance valuesfor our 233 friend lists and we found a significant positive bias. Over 48% of the friends lists have values larger than 0.2, suggesting that a large fraction of friend lists could be automatically inferred from the social network.
การแปล กรุณารอสักครู่..
