Abstract Based on the BERT model to analyze online public opinion on social media, the public opinion information is transformed into an intuitionistic fuzzy evaluation matrix, and the temporal change of public opinion is analyzed to calculate each temporal weight, and using the generative large language model to simulate the decision makers’ behaviors, and the news text and online public opinion are provided to the large language model to simulate the opinions of decision makers with different risk preferences, so that we can propose a multi-stage public opinion-driven group consensus decision-making method. By combining the degree of hesitation and the degree of trust, we calculate the maximization adjustment degree of decision makers, and considering the individual differences of decision makers, give the calculation methods of attribute weights and consensus thresholds, and construct a dual-path feedback model to address conflicts among decision makers. The feasibility and effectiveness of the method in group decision-making are verified through a case study involving investment selection of new energy automobile enterprises.
LIU Min,ZHANG Luxiang,PING Weiying等. Research on Multi-Stage Online Public Opinion-Driven Group Consensus Decision-Making Method Based on Large Language Model[J]. Chinese Journal of Management, 2025, 22(4): 750-.
