Abstract: In large group decision making, the opinions of every expert are very important, and minority opinions are also very important, but there are risks in considering minority opinions. This paper proposes a risky large group decision-making method for protecting minority opinions in the social network environment, to deal with minority opinions that may exit and to improve the quality of decision-making. Firstly, the natural language processing technology is used to extract the keywords of the public information about the incidents in social media, and then TF-IDF technology is used for big data processing to extract the event attributes of public interest and determine the weights of attributes. Secondly, the relationship between experts is taken into consideration in decision-making. Then, the identification and risk measurement mechanism of minority opinions are established, and two risk factors are defined: trust risk coefficient and preference risk coefficient. We adjust the weights of minority opinion groups through the trust risk coefficient of minority opinions and regulate group consensus through preference risk coefficient, to obtain a large group decision-making plan with higher level of consensus; Finally, through the case analysis of “7.1” flood in Ningxiang, the feasibility and effectiveness of the proposed method are verified.

Key words: social network, minority opinions, large group, risk, emergency decision-making

摘要： 大群体决策过程中,少数意见十分重要,但考虑少数意见又有一定风险性,本文提出了一种社会网络环境下保护少数意见的风险性大群体应急决策方法。首先,利用自然语言处理技术对社交媒体中公众对突发事件的传播信息进行关键词提取,再采用TF-IDF技术提取公众关心的事件属性并确定属性权重;其次,将专家之间的联系考虑到决策当中,构建决策专家之间的社会网络关系;然后,建立少数意见的识别和风险测度机制,并定义了两个风险系数:信任风险系数和偏好风险系数,通过少数意见聚集的信任风险系数调节少数意见聚集的权重,通过偏好风险系数调节群体共识,以得到共识水平较高的大群体决策方案;最后,通过“7.1”宁乡洪灾案例分析,验证了所提方法的可行性和有效性。

关键词: 社会网络, 少数意见, 大群体, 风险, 应急决策