The Two‐Stage Hierarchical Consensus Reaching of the Large‐Scale Group Decision‐Making Based on Social Network

ABSTRACT In real‐world scenarios, group decision‐king (GDM) is essential due to the involvement of multiple stakeholders, the integration of complex information, and the need for equitable solutions. Addressing the inherent challenges in reaching consensus among diverse decision‐makers necessitates the development of sophisticated models. This study proposes a two‐stage multi‐level large‐scale group consensus model. First, high‐cohesion expert subgroups are formed by constructing a similarity matrix via kernel functions and applying spectral clustering. Second, to accelerate consensus, a hierarchical feedback mechanism is introduced. Experts are categorized into three tiers based on the subgroup's consensus level, each triggering a distinct adjustment strategy. For low consensus groups, a feedback adjustment is established based on individual opinions, aiming for minimal individual adjustments and group conflict, with the particle swarm optimization (PSO) optimization rules used to address feedback suggestions. For moderate consensus groups, expert preferences deviating from the group opinion are adjusted based on their confidence levels and group opinions. To obtain comprehensive and objective attribute weights, this study combines entropy weighting methods with matrix decomposition. Finally, the effectiveness of the proposed method and its superior clustering performance are demonstrated through case analysis and comparative analyses.