The spontaneous behavior in extreme events: a clustering-based quantitative analysis

Social media records the pulse of social discourse and drives human behaviors in temporal and spatial dimensions, as well as the structural characteristics. These online contexts give us an opportunity to understand social perceptions of people in the context of certain events, and can help us improve disaster relief. Taking Twitter as data source, this paper quantitatively measures exogenous and endogenous social influences on collective behaviors in different events based on standard fluctuation scaling method. Different from existing studies utilizing manual keywords to denote events, we apply a clustering-based event analysis to identify the core event and its related episodes in a hashtag network. The statistical results show that exogenous factors drive the amount of information about an event and the endogenous factors play a major role in the propagation of hashtags.