The rapid growth of the digital economy has significantly enhanced the convenience of information transmission while reducing its costs. As a result, the participation in social networks (SNs) has surged, intensifying the mutual influence among network participants. To support objective decision-making and gather public opinions within SNs, the research on the consensus-reaching process (CRP) has become increasingly important. However, CRP faces three key challenges: first, as the number of decision-makers (DMs) increases, the efficiency of reaching consensus declines; second, minority opinions and non-cooperative behaviors affect decision outcomes; and third, the relationships among DMs complicate opinion adjustments. To address these challenges, this paper introduces an enhanced CRP mechanism. Initially, the hippopotamus optimization algorithm (HOA) is applied to update the initial community division in Leiden clustering, which accelerates the clustering process, collectively referred to as HOAL. Subsequently, a two-stage opinion adjustment method is proposed, combining minority opinion handling (MOH), non-cooperative behavior management, and dual-fine tuning (DFT) management, collectively referred to as DFT-MOH. Moreover, trust relationships between DMs are directly integrated into both the clustering and opinion management processes, resulting in the HOAL-DFT-MOH framework. The proposed method proceeds by three main steps: (1) First, the HOAL clusters DMs. (2) Then, in the initial CRP stage, DFT manages subgroup opinions with a weighted average to synthesize subgroup perspectives; and in the second stage, MOH addresses minority opinions, a non-cooperative mechanism manages uncooperative behaviors, and DFT is used when negative behaviors are absent. (3) Third, the prospect-regret theory is applied to rank decision alternatives. Finally, the approach is applied to case analyses across three different scenarios, while comparative experiments with other clustering and CRP methods highlight its superior performance.
