The　surge　of　social　media　has　made　large-scale　group　decision-making　(LSGDM)　crucial　in　real-world　decision-making.　The　intricacies　of　trust　relationships　within　social　networks　that　emerged　from　relations　in　social　media　affect　both　the　clustering　and　the　consensus　of　large　groups.　However,　existing　research　often　neglects　the　impact　of　overlapping　social　trust　networks　on　group　consensus.　To　fill　this　gap,　this　study　introduces　a　novel　consensus-reaching　process　(CRP)　that　integrates　overlapping　community　detection　and　ELICIT-based　optimization　models　under　bounded　confidence.　Initially,　the　Lancichinetti-Fortunato　method　(LFM)　is　employed　to　identify　overlapping　community　structures　within　social　trust　networks,　delineating　several　subgroups　and　identifying　corresponding　non-overlapping　and　overlapping　decision-makers　(DMs).　Subsequently,　the　PageRank　(PR)　algorithm　is　utilized　to　compute　both　global　and　local　weights　for　individuals,　facilitating　a　rational　aggregation　of　collective　and　subgroup　opinions.　Next,　two-stage　Extended　Comparative　Linguistic　Expressions　With　Symbolic　Translation　(ELICIT)-based　optimization　consensus　models　under　bounded　confidence　are　designed,　aiming　to　provide　optimal　feedback　for　guiding　DMs＇　preference　adjustments.　Since　overlapping　DMs　may　belong　to　multiple　subgroups,　a　weighted　influence　feedback　mechanism　is　introduced　to　mitigate　conflicting　guidance　from　these　multiple　affiliations.　Finally,　we　demonstrate　the　effectiveness　and　superiority　of　our　proposed　method　through　numerical　validation　and　comparative　analysis　against　existing　approaches.　©　2024　Elsevier　Inc.