论文部分内容阅读
突发事件所引发的群体情绪可能会导致衍生群体事件的发生,所以有必要对群体情绪演化进行监控预警。本文对突发事件发生、发展过程中群体情绪的演化进行了研究,建立了面向突发事件的情绪层次模型以及情绪监控预警模型。通过支持向量机、贝叶斯文本分类算法、n元语言模型三种机器学习算法,以及CH I(卡方检验)、DF(文档频率)、IG(信息增益)三种特征选择方法对突发事件相关的微博评论进行了情绪分类的实验分析,实验结果表明支持向量机、IG取得的分类性能最优,得到的分类结果可以满足实际应用的需求。进而对王家岭矿难、山西疫苗两个突发事件进行了实际案例研究,建立并分析了情绪层次演化曲线。最后提出了群体情绪演化预警指标以及预警模式的概念。
The group emotions triggered by unexpected events may lead to the occurrence of derivative group events, so it is necessary to monitor and alert group emotional evolution. In this paper, the evolvement of group emotion in the process of emergencies and development is studied, and the emotion level model for emergency and the early warning model of emotion monitoring are established. Three kinds of machine learning algorithms, such as support vector machine, Bayesian text classification algorithm and n-gram language model, as well as three feature selection methods of CHI (chi-square test), DF (document frequency) and IG (information gain) The related micro-blog comments conducted an experimental analysis of emotion classification. The experimental results show that the classification performance obtained by IGS is the best, and the classification results obtained by IG can meet the needs of practical application. Then the case study of Wangjialing mine accident and Shanxi vaccine two emergency cases was carried out, and the evolution curve of emotion level was established and analyzed. Finally, the concept of early warning index and warning mode of group emotional evolution is put forward.