Research On The Information Forwarding Prediction Model Based On Weibo

Massive user groups have caused the online social network represented by Weibo to have an increasingly large social influence,and played a great role in the spread of many social hotspot information and major public opinion events.At present,due to the fact that online social networks are much more complicated and the factors that affect communication are more abundant,the society still has insufficient understanding of the core regular pattern of information dissemination on online social networks.In-depth research on information dissemination patterns and prediction of communication behaviors and trends have become urgent and hot issues.The main work of this article is aimed at the propagation prediction of the main mechanism of the microblog platform,the purpose is to study the trend and scale of the microblog information spread in the network,and to predict the popularity of the information.It is mainly developed from two aspects: first,based on the infectious disease model and the cascade model,to study the internal laws of microblog propagation,establish a regression model,and predict the amount of microblog reposting;in addition,analyze the factors that affect the reposting of microblogging,and generate feature spaces,Use machine learning algorithms to classify and predict the forwarding volume of Weibo.The main contents are as follows:1.Establish a regression equation to predict the forwarding volume of Weibo.First,combining the infectious disease modeling theory,the Weibo users are divided into two compartments,and the fans of the communicator are potential followers of the Weibo,establishing a quantitative relationship equation between Weibo followers and transferers,and then according to the two the state transition pattern between the two warehouses,the establishment of a predictor model of the number of communicators,and finally through the quasi-Newton method,using the forwarding data of the initial propagation of Weibo to carry out experiments on the parameters in the model on the data set and analyze various parameters The error under the setting is compared with the benchmark model to verify the validity and reliability of the model.2.Utilize the initial forwarding data and user behavior data of Weibo to establish a classification model and predict the forwarding scale of Weibo.First of all,the characteristics of dating,and the forwarding network formed during the propagation of Weibo is counted,thus expanding the feature space of Weibo forwarding.Then use theBi-GRU neural network to extract the features of the microblog text,thereby improving the accuracy of the semantic feature description,and integrating with the publishing features of the microblog to establish a more comprehensive feature engineering.After normalizing the features,input to the step-up training(based on GBDT)for classification.Finally,a microblog popularity prediction model based on multi-dimensional features is obtained through training.Experiments show that,with the same type of association,the model can accurately identify the popularity of Weibo.Therefore,an excellent analysis of the features shows that the temporal features added here and the improved semantic features can effectively improve the prediction effect.