Research And Implementation Of Trend Analysis For Hot Events Of Public Opinion In Social Media

With the improvement of network facilities and the popularization of mobile Internet,people can get and release information through social media more quickly.Social media has become an important channel for daily communication.To find and analyze the hot events in social media is helpful for researchers to understand the development trend of events and provide data reference for public opinion intervention and guidance.Compared with news articles,social media content has the characteristics of strong colloquialism,short text length and more useless text,which makes news event recognition technology and analysis method can not be directly applied.At the same time,event analysis is mainly based on statistical characteristics.How to import text semantic information to improve the accuracy of event discovery and analysis in social media needs further research.In view of the above problems,this paper analyzes the trend of events based on the semantic features of the text.This paper proposes a hot event recognition model based on multi-dimensional features,which combines statistical features and text semantic features,integrates clustering methods,trains event classification model,and discovers network hot events.This paper also proposes a multi-dimensional event trend analysis method,which analyzes the event evolution based on the accumulation of event keywords in the time dimension,analyzes the influence of event popularity,text originality,user participation and content sensitivity on event popularity measurement,and trains neural network to predict event trends.This paper also designs and implements a hot event analysis system based on proposed models.The experiment obtains the data of University Forum,Sina News and Sina Weibo,and validates the model from three aspects:event recognition effect,event evolution analysis,and event popularity prediction effect.Compared with the model based on statistics features,the F value of the event recognition model proposed in this paper is increased by 24.6%and 5.7%compared with other models based on semantic features;the event evolution analysis method proposed in this paper can discover the event changes in time and extract the event keywords that are easy to understand;the accuracy of the event heat prediction model proposed in this paper is increased by 9%and 10%respectively compared with two reference models.The experimental results show that the event detection and analysis method based on text semantic information proposed in this paper can discover hot events in social media more efficiently,and analyze the development trend of events in detail.