Research On Detection And Prediction Technology Of Hot Topics In Emergency Internet Public Opinion

As one of the main battlefields of Internet public opinion,it is of great significance to study Internet public opinion in Weibo.Most of the data in Weibo is about peoples' daily lives,and only some of them contain information about popular events.Therefore,in the face of complex and diverse microblog information,how to detect hot topics accurately and make predictions effectively becomes more and more important.However,the current researches didn't take the complexity of microblog hot topic metrics and its transmission in social networks into account,resulting in low detection and prediction accuracy of hot topics,especially for sudden hot topics.For this reason,this thesis solves the above problems by introducing multi-dimensional hot topic measurement model and optimized random forest hot topic prediction method,and carries out corresponding research.The research results in this thesis are followings:1.A multi-dimensional hot topic measurement model is proposed for the problem of low detection accuracy of sudden hot topics.Firstly,this model filters these topics and obtains a hot topic initial set,then integrates the influence factors of the topic heat,and calculates the comprehensive weight of each topic.And then,the comprehensive weight of the topic is effectively combined with the multi-dimensional hot topic measurement model according to a certain weight,and finally a hot topic detection model based on microblog multi-dimensional and comprehensive weight is obtained.The experimental results show that the proposed algorithm model in the detection of sudden hot topics has a higher detection accuracy than the traditional algorithm,and the overall performance is stable,thus improving the quality of sudden hot topic detection.2.In order to improve the prediction accuracy of sudden hot topic,the optimized random forest prediction method is used in this thesis to predict the hot topic of Weibo.The method selects the feature according to the weight of different features.The more the feature influence the weight of the heat,the easier it is to be extracted so that it can improve the accuracy of the prediction.Finally,compared with two traditional algorithms,the experimental results show that the prediction effect of the optimized random forest prediction method is significantly improved compared with the other two algorithms.3.This thesis uses Java Web Framework to design and implement the prototype system of the above model,and uses Web pages to display personalized detection and prediction results for hot topic detection and prediction users.