Community　discovery　can　help　reveal　the　topological　structures　and　the　dynamic　characteristics　of　the　real　social　networks.　However,　most　community　discovery　algorithms　are　designed　for　static　social　networks,　while　in　most　real　social　networks,　the　community　structures　of　networks　change　dynamically.　For　community　discovery　in　dynamic　social　networks,　existing　algorithms　are　based　on　the　assumption　that　the　community　structures　of　dynamic　networks　evolve　smoothly,　and　therefore　cannot　deal　with　the　sudden　events　of　emergence　and　extinction　of　communities　during　the　evolution　of　dynamic　social　networks.　To　address　the　aforementioned　issue　of　discovering　community　structures　effectively　and　efficiently　in　large-scale　dynamic　social　networks,　this　paper　presents　a　novel　community　representation　model　called　Follow-Community　model,　which　represents　the　following　relationships　among　neighbors.　In　the　Follow-Community　model,　the　community　is　represented　by　nodes　with　different　roles　and　the　following　relationships　among　the　nodes.　By　finding　the　direct　or　indirect　following　relationships　among　nodes,　the　collection　of　nodes　that　follow　the　same　leader　can　be　partitioned　into　one　community.　Based　on　the　Follow-Community　model,　a　Neighborhood　Following　Algorithm　(NFA)　with　nearly　linear　time　complexity　is　proposed　to　discover　community　structures　in　static　social　networks　by　just　traversing　the　node　set　only　once.　Furthermore,　an　extended　algorithm　of　NFA,　named　incremental　Neighborhood　Following　Algorithm　(iNFA),　is　also　proposed.　By　updating　the　neighborhood　following　relationships　of　the　relevant　nodes　involved　in　network　evolution　over　time,　iNFA　can　discover　the　community　structures　and　community　evolution　in　dynamic　social　networks.　Finally,　extensive　experiments　are　conducted　to　validate　the　advantage　of　the　proposed　algorithms　in　accuracy,　efficiency　and　stability　for　community　discovery　in　large-scale　dynamic　social　networks.　©　2017,　Science　Press.　All　right　reserved.