Research On Real-time Monitoring And Early Diffusion Prediction Model Of Microblog Events

With the rapid development of social networks, microblogquickly became one of the main information media andbirthplaces of the networkeventsbased on its ease, extensive, real-time and low-cost, and results in important social influence. It has a positive meaning for the control and guidance network event to study the diffusion mechanism and potential law of tweets, and the early prediction of proliferation scale.First Page Rank,an evaluation algorithmof influence for microblog users, and its improved algorithm UIR(user influence rank) are described. Then a new calculation method is put forwards, considering multiple factors on three aspects, breadth of microblog audience, contents and information dissemination of microblog. The sample data of influential events arecollected to verify the reliability of the method.Previous studies for the diffusion law of microblog network eventsare summarized. Then feature data of network events diffusion are extracted and analyzed from big data resources, and general rules of the early spread of the microblogging network eventsare proposed and validated. They are active- sleep patterns corresponding to schedule, geographical proximity associated with incident region, timeliness of the event spread consistent with the characteristics of microblog,short distancepatternsclosely related tosocial network structure, and strength of subsequent forwarding fitting with propagation mechanismof microblog.The propagation rules and diffusion characteristicsof microblog information are given. They are one-to-many dissemination with several information centers, and explosive spread based on network structure. Feature information extractedfrom network events is computed in chronological orderto form time series and establish dynamic BP neural network model to predict diffusion range of network event. The dynamic BP neural network is trained and tested by early monitoring data of network events. The predicted error of diffusion range is less than 10%, and it proves the validity of the model.The main innovations of this paper are summarized as follows:A new calculation method is put forwards to evaluatethe influence of microblog users, considering multiple factors on three aspects, breadth of microblog audience, depth of contents and information dissemination of microblog. The sample data of influential events are collected to verify the reliability of the method.Various factors are real, easy to get, and simple to calculate.Dynamic BP neural network model is established to predict diffusion range of network event by monitoring and extract feature information from network events. The spread of network events is predicted to timely control and guide the network events.