Dynamic Study Of Micro-blog Public Opinion Under Emergency Events Situations

Recently,the social media has become a social sensor for emergency event.The online public opinion on it can reflect the development of the emergency quickly and directly.Micro-blog as an important channel of public opinion,can make tremendous impact on the community after the experience of emergency.Public can form and spread their views,attitudes and emotions quickly.The accurate early warning of public opinion must be done timely.If not,there may trigger a new burst of online events,even the next group event.Therefore,the pre-warning of public opinion for emergency situations is important.However,existing research has achieved fruitful results,there are still some problems.Due to most of the current researches separated the Micro-blog text contents,emotions and the information behavior,the interplay among these three dimension can't be obtained.So,the early-warning model is lack of theoretical supports.In addition,some of the current studies are static snapshot warning,it is difficult to satisfy the real-time monitoring of public opinion on development trend of early-warning requirements.Aiming the problems above,this paper makes some efforts.Firstly,put forward a method to identify the dominant emotion based on the causal bayesian network.Obtained the interplay rule between micro-blog emotion and the users' information behavior by the sensitivity analysis and regression analysis.Secondly,study the relationship of micro-blog text contents and micro-blog emotion.Based on LDA model and knowledge map to identify the feature words of different stage.And then analyze the interplay rule between micro-blog text contents and micro-blog emotion used Lasso method.The rules obtained above can be used to give theoretical supports to the construction of pre-warning model.Based on the rules to construct the warning index,and then construct the pre-warning bayesian network model.Finally,collecting large number of micro-blog data about the emergency event,to validate the validity of the model.This paper considers the interplay among micro-blog text contents,micro-blog emotions and users' information behaviors fully,which provides good theoretical supports to the construction of warning index and the initial structure of pre-warning model.The pre-warning model for emergency micro-blog public opinion based on bayesian network can provide anaccurate warn to the public opinion timely,which can give the managers technology and decision supports.