In　recent　years,　large-scale　group　decision-making　(LSGDM)　has　garnered　significant　scholarly　attention.　Given　that　decision-makers　(DMs)　may　come　from　various　departments　and　have　distinct　knowledge　backgrounds,　they　often　use　heterogeneous　information　to　express　their　assessments.　To　address　these　challenges,　this　paper　proposes　a　new　decision-making　model　for　large　groups　in　a　social　network　environment.　Initially,　the　large　group　is　divided　into　communities　based　on　the　trust　relationships　among　DMs　using　community　detection　algorithm.　Subsequently,　a　direct　method　is　used　to　process　DMs’　heterogeneous　information.　Simultaneously,　we　integrate　the　weight　information　of　DMs　and　communities　to　calculate　the　preferences　of　the　group　using　a　weighted　averaging　operator.　Additionally,　an　innovative　feedback　mechanism　is　designed,　which　takes　into　account　the　bounded　confidence　of　experts　and　the　influence　of　community　leaders,　to　enhance　the　consensus　level.　Finally,　the　feasibility　and　effectiveness　of　the　proposed　model　are　demonstrated　through　a　specific　case　study.　©　2025　Northeastern　University,　China.