Research On Event Prediction Of Social Network Events Based On Text Analysis

With the rapid development of the Internet,social networks,as the main platform for information exchange,have played an increasingly important role in people's lives.The explosive growth of information makes the heat prediction of event in social networking platforms a challenging research issue.Previous researchers have focused their research on propagation heat prediction on network structure or surface feature prediction algorithms.They have not adequately analyzed the text content of social network events and neglected the key information in text semantics.This article will focus on the heat features included in the content of social networking event texts,and extend the surface features of traditional prediction algorithms to improve the accuracy of predicting.The traditional heat prediction method considers the event's likes,comments,forwarding numbers,and whether the event's main sticker contains surface heat features such as pictures,videos,links,and special characters.This paper analyzes the text content in the network event,extracts the semantic information in the text content,expands the surface heat feature of the social network event,and thus improves the prediction accuracy.Specifically,it includes extracting event subject influence characteristics,text category characteristics,and event level characteristics in the event text.First,we use the HMM named entity recognition method to extract the main set of events according to the main text content of the event,and calculate the subject's influence as the subject's influence characteristics of the event.Afterwards,by establishing a social network event rank system,the event rank features are comprehensively obtained by using text classification and the current propagation heat of the event.Finally,after synthesizing surface features,the propagation heat feature vector of the event is constructed.And use the high-performance GBDT regression prediction algorithm to model and predict the temperature of social network events.Through experiments,according to the principle of control variables,after comparing the RMSE,MAPE,and PCCs evaluations of the single surface feature and extended feature training model,it is proved that the extended predictive feature model can effectively improve the prediction accuracy of social network events.Based on the above,this paper designs and implements a heat prediction system that includes three social networking platforms: Sina Weibo,Twitter,and Facebook.It has achieved good performance in practical applications.