Knowledge graph clustering and dual strategies for adaptive mechanism in large-scale group decision making

Clustering and feedback mechanisms are important methods in the large-scale group consensus framework. This article proposes a novel clustering method rooted in objective information similarity that uses knowledge graph technology along with the development of a dual-strategy-driven comprehensive adaptive feedback mechanism (CAFM). An expert knowledge graph is constructed using objective textual information to extract relationships that are converted into semantic vectors using Knowledge Graph Embedding (KGE). This approach can quantify experts’ similarities and their subsequent clustering. A reliability degree index for subgroups is defined to quantify willingness to adjust preferences in conjunction with their sizes and is proposed to derive subgroup weighting values. Building on these foundations, the CAFM is developed to guide opinion interactions. Feedback coefficients are generated using subgroup reliability and similarity values to determine adjustment willingness levels. Two optimization functions are constructed to address diverse strategic needs: a single-objective function that minimizes adjustments and a bi-objective function that simultaneously minimizes adjustments and maximizes satisfaction. Finally, the proposed model’s applicability and effectiveness are validated via a case study involving the selection of small and medium-sized enterprises (SMEs). Results underscore the potential of the model to improve consensus efficiency and decision making quality in large-scale group decision contexts.