To solve the problem that it's difficult to obtain the information of decision attributes in the complex large group emergency decision under the social network environment, a large group emergency decision-making method driven by big data of public behavior under the social network environment is proposed. Firstly, by mining the big data of public behavior on social platforms, the TF-IDF and Word2vec technology are used for the extraction, clustering and influence analysis of Keywords. Large group decision attributes and their weights are mined from a large amount of behavioral data to assist expert decision making, so as to make the decision results more scientific and effective. Secondly, a social network based on trust relationship and similarity of views among decision makers is constructed. The decision makers are clustered by a clustering method considering both trust and similarity, and the weights of decision makers are obtained based on social network analysis. Then, a consensus adjustment method considering the relationship of trust among decision makers is proposed to obtain the final group decision matrix and alternatives ranking. The objective confidence is introduced to avoid the influence of overconfidence behavior of individual decision makers. Finally, a case study about coronavirus is given to illustrate the feasibility and effectiveness of the proposed method.