LDA-DeepHawkes Model

With the vigorous development of the Internet,the emergence of various new online social media provides a new way for the spread of information,which greatly promotes the generation and transmission of information.The spread of predictive information on social media has important application value for information recommendation,rumor control,public security,trend topic tracking and other aspects,and is an important research direction of social network analysis.With the rapid development of the Internet and the emergence of various new online social media,new ways of information dissemination and diffusion have been provided,which greatly promote the generation and transmission of information.The spread of predictive information on social media has important application value for information recommendation,rumor control,public security,trend topic tracking and other aspects,and is an important research direction of social network analysis.The existing prediction methods of information popularity are mainly divided into feature-based method,generation method and the combination of the two methods mentioned above.At present,DeepHawkes model has a better prediction effect,but it only models the early diffusion of information cascade to predict the future popularity of information,ignoring the influence of the text content of information on the popularity of information transmission.But in fact,the text content of information plays an important role in the popularity of information dissemination,and there is a strong correlation between them.Based on DeepHawkes model,lda-deephawkes model considering both cascade factors and text content is proposed.The main work of this paper includes:(1)DeepHawkes model considering topics was proposed.LDA topic classification model was used to extract the content of message text topics,and deep model was used to learn the presentation vector of topics to simulate the self-excitation mechanism between topics.(2)the lda-deephawkes model is proposed,which integrates the DeepHawkes model that only considers information cascade with the DeepHawkes model that only considers subject,inherits the high explanatory power of the DeepHawkes model,and further improves the accuracy of prediction results(3)sufficient experiments were carried out on two real data sets,various detailed parameters in the model were optimized,and the results were compared with the predicted results of other models,verifying the effectiveness of the model proposed in this paper.