Research On Emergencies Detection And Key-Nodes Mining

Ibday, the rapid development of the Internet, information is disseminating at an unprecedented pace and breadth. Four major online social media - micro-blog, social networking sites, open encyclopedia, video sharing sites are affecting people's production and life all the time. Among them, micro-blog become an important information dissemination and access platform because of its real-time and convenience features. In the face of the complex and diverse micro-blog data, people need a kind of technology to quickly get valuable information, especially the recent hot events. As for public opinion supervision department, only these information is not enough, also need to further explore the development and dissemination of the emergency public opinion, timely control the important node which spread negative emotions in the network, in order to achieve the purpose of public opinion supervision.Contents of this paper are mainly the following aspects:(1) A new micro-blog emergencies detection approach base on the H-index of was proposed. Based on the emergency of words and their cover degree on documents, the concept of H-index of burst words was proposed which is used to extract the features of the emergent events. By constructing the burst text-document matrix, combined with the classic hierarchical clustering algorithm to cluster the micro-blog text. Based on the original micro-blog in each cluster, the emergencies were determined. Experiments on Micro-blog Sina data set show that the approach proposed in this paper could effectively identify emergencies.(2) A new approach to mining key nodes from emergency information dissemination network combined with entiment analysis was proposed. Construct a information dissemination network for a special emergency. Basing on the activity of nodes and their ability to control information flow, the key-nodes which have larger influence on information flow were extracted. Then the sentiment analysis was used to detect the key-nodes with negative emotions. These nodes play key roles in public opinion direction, which could provide important information for public opinion supervision.In summary, this paper detect emergencies in micro-blog data first, then dig out the key nodes in emergency information dissemination network combined with entiment analysis, identifies key nodes with emotional tendencies, thus providing a strong basis for public opinion supervision.