Research On Weibo Propagation Prediction Based On Triplet Neural Network

With the rapid development of Internet technology and the continuous enhancement of the functions of online social platforms,many users choose to use Weibo as an online social platform,and the spread of positive or negative Weibo content has greatly affected people's lives.The dissemination speed and scope of Weibo are affected by the features of Weibo content and the features of users who publish Weibo.Research on Weibo dissemination prediction is of great significance in promoting positive energy,commodity marketing,public opinion guidance,and blocking the spread of bad information.This paper analyzes the features of Weibo from three aspects: the text features of Weibo(text content of Weibo and topics and other features),the user influence features of Weibo(whether the Weibo user is authenticated and the number of followers of the Weibo user)and the multimedia features of Weibo(whether the Weibo contains the features of music,pictures and videos),and proposes a multi-feature Weibo propagation prediction model based on triplet neural network to improve the accuracy of Weibo propagation prediction.The main work of the paper includes two aspects:(1)Feature analysis and extraction of multimedia features of Weibo and propagation heat prediction of multimedia features of Weibo by using neural network.In this paper,convolutional neural network is used to extract music features of Weibo,SIFT algorithm is used to analyze and extract features of Weibo pictures,and PEARL model is used to extract video features of Weibo.Triplet neural network is used to train and classify the extracted microblog multimedia features,identify the high popularity multimedia features and low popularity multimedia features and mark the prediction results.(2)A prediction model of multi-feature microblog propagation based on threebranch neural network is proposed.Using the LDA topic model to extract features from the text content of Weibo,using the improved PageRank algorithm to extract features of the influence of Weibo users,combining these features with the multimedia features after the heat mark,and input it into the triplet neural network for training,so that the triplet neural network can predict whether a Weibo will become a highly popular Weibo based on the text features,user influence features and multimedia features of the Weibo.Compared with other three Weibo propagation prediction models,the experimental results show that the multi-feature Weibo propagation prediction model proposed in this paper based on triplet neural network has improved prediction accuracy compared with the original model,and the model is more stable than the original model.The performance of the model in this paper is good,and it has certain reference value and exploration significance in the research field of Weibo propagation prediction.