Research On Emotional Abnormal Detection Based On Weibo Review Data

In the face of today's information age,the rapid development of the Internet requires the protection of corresponding management system.Timely collection and analysis of netizens' emotions and mentality is an important prerequisite for the government to make correct decisions and solve social conflicts.In the research field of netizens' emotions in emergencies,few foreign scholars have been involved in it,and few studies have taken netizens' emotions as the subject alone.Most of them only mention user emotions in the research of users' behaviors on social networking sites.Domestic scholars have studied netizens' emotions to some extent,but they still remain at the level of qualitative analysis.They simply enumerate the government's response measures to netizens' negative emotions,as well as the use scenarios,but lack relevant studies that discuss the government's emergency response strategies based on actual events and data.Therefore,it is necessary for us to discuss and study netizens' emotions in emergencies.Firstly,the techniques and methods related to subject research are introduced:feature extraction technology,text classification method,cluster analysis method,cluster-based anomaly detection method,fuzzy theory,emergency decision theory,case-based reasoning method,content-based algorithm is recommended and the application scenarios of these theories and methods are analyzed.Secondly,based on the "Knowledge Network Emotion Dictionary" and "Harbin Institute of Emotion Dictionary" as the basic emotion dictionary,the SO-PMI technique is used to extract the emotional vocabulary containing the network characteristics from the microblog corpus,and the basic emotion dictionary is combined with the characteristic emotion vocabulary.The method determines the emotional dictionary of the subject,and proposes the use of the backpack model to construct the anomaly feature set in the mathematical theory of the knapsack problem.Then based on the characteristics of the fuzzy clustering algorithm,an improved fuzzyclustering algorithm is proposed to effectively prevent the local optimality.In addition,according to the emergencies and the emotional characteristics of netizens,f this paper describes the emergency methods for netizens' abnormal emotions in the face of emergencies from the aspects of case retrieval and matching,case correction and preservation.Finally,to aim at “2016 Wuhan Flood Disaster” case,the improved sentiment dictionary is used to calculate the emotional value of the comment text,the abnormal standard is formulated according to the sentiment analysis result,and the abnormal review text matrix vector is constructed by using the classic backpack model.The event netizen comments text,constructs an anomalous text feature set,and based on this feature set,uses the improved C-means clustering algorithm as a method to achieve abnormal emotion detection.The experimental results show that,in order to accurately grasp the emotional tendency of user comments,taking the emergency as the background,it is proposed that the abnormal detection scheme of comment data,which is based on the emotion acquisition,emotional feature extraction and abnormal emotion detection is complete and technically feasible,and has important practical significance for the control and management of Internet users' emotions.