Public Opinion Events Active Detection Research On Microblogging

With the rapid development of Internet in China, our country is in the strengthening and innovation of social management. It’s an important period for improving public opinion guiding mechanism on Internet. At the same time, along with the increasing socialization of information network, the representative of the new media to microblogging has become the main source of the outbreak and spread of network public opinion events. Especially, for the past few years, in the event of process of occurrence and development, microblogging which spreading based on social relationships, has become an important driving force for the network public opinion events. To explore how to use the Internet to collect public opinion to serve the people’s livelihood better is a hot topic in the research of information management and technology. But the method and technology of establishment of microblogging public opinion information system is still not mature enough in domestic and foreign in the current. The difficulty lies in the uncertainty, complexity, the widely subject of network public opinion events, and microblogging itself has short text, semi-structured, complex structure features of social networking, etc. That makes the traditional the detection method of public opinion hot events based on words is not applicable, and construct a method that adapted to the characteristics of microblogging to detect network public opinion events has become an urgent problem to solve.This paper puts forward a method of microblogging public opinion events active detection which access to information on various dimension broadly through the microblogging web crawler, and then analyses the various dimensions of public opinion respectively. The last, calculate the composite dimensional information features of microblogging public opinion events through the text characteristics and social characteristics of microblogging for unsupervised clustering. Thus, realize the microblogging network public opinion active perception of the event.There are two innovation points in this paper:(1) Established the content and social relations subject characteristics of the microblogging network events mining model based on LDA. In the work of early warning and disposal of the network events, the theme is difficult to define, which includes not only the content features of microblogging, but also social relationship characteristics of microblogging need to deal with, and only fully tap the content and relationship between the characteristics of microblogging network public opinion events, can we identify potential microblogging network public opinion for maximum.(2) Put forward an incremental SVM pattern recognition algorithm for microblogging public opinion events dynamic perception. Microblogging network events of public opinion are strongly associated with the dimension of time. In different periods, the results will be completely different. According to the characteristics of the microblogging network public opinion events of the dynamic and constantly changing in a certain period of time, incremental SVM algorithm will be used in this paper to classify new microblogging that related with incremental pattern recognition, and than calculate the changing public opinion of hot spots over timeline.Simulation experiments of microblogging public opinion events perception of prototype system show that the method proposed in this paper for microblogging topic mining and dynamic tracking on public opinion have a good effect.