Research On The Survival Analysis And Prediction Of Online Social Network Topics

With the generation of Web2.0technology, the way human beings use the Internet has a profound transformation, from simple information searching and web browsing to online social relationships construction and maintenance. The emergency and rapid development of microblog, blog, forums, wikis and other social network applications are changing the way human beings’ communication. Online social networks integrate the characteristics of high interactivity and high community participation. The integration of numerous content and point of views makes the generation of influential topics to be possible. These influential topics attract Internet users’ attention and identify clear opinions, which are the origins of online public opinions and play an important role on the formation of social point of view. Therefore, it is essential to research and predict the developing trend of online social topics.Currently, the related analysis on the survival of online social network topics is still at an early stage. For the deficiency of this field, this study uses large amount of objective data on Sina microblog and online social network group discussing topics as study case and study subject to explore the survival of social network topics from three aspects respectively:social network topic influence, topic persistence and topic directionality. And related results are shown as follows:1. Proposes a new concept Microblog topic influence index (topic-I index) which is composed from three dimensions:participants, topics and network. Applying PCA and huge volume of real data to assign each index weight, we eventually obtain7principal components and the calculation methodology of topic-Ⅰ index, which aims to provide network public opinion decision supports for topic development trend prediction. 2. Establishes a survival analysis model to research and predicts the influential factors of topic persistence and the corresponding influence degree. Finally we found that the most influential factors are topic picture information volume, Internet users’attention degree and topic administrator’s visibility respectively. Moreover, we also divide these topics into different risk groups by calculating the prognostic index and expected survival rate of them. Finally classify these topics into4categories according to the scientific criterion and make regression analysis to explore which factors influence most. This help predict and control online topics’development trend according to time dimension.3. Combines the mechanism of group polarization, makes prediction on opinion’s development direction of online topics. This study found that the most influential factors affect the development direction are group homogeneity, topic sensitivity, topic ambiguity and the first affected Internet users. This provides great value for enterprises to grasp consumers’trends and government to seize public opinions spreading directions, which in order to incite topics’positive transmission and intervene public opinions’negative spreading.This research provides better topic managerial insights for topic administrators, and further gives reference guide and positive meaning to spread and control online information.