Rumor Identification Based On Micro-blog

Micro-blog is one of the most popular social platforms in China,which allows users to quickly obtain information in various aspects.The characteristics of micro-blog,such as timeliness and diversification,have attracted many netizens,but at the same time,the rumors of micro-blog have spread rapidly through micro-blog.Rumors can spread in a short period of time and have a wide range,resulting in great risks in various aspects,so how to effectively identify false information and reduce the harm caused by rumors and public opinion to the masses is of great research value.This thesis takes micro-blog rumors as the research object,extracting features from the three basic aspects of micro-blog users,mode of transmission,and text information,and also takes into account the deep characteristics micro-blog user reviews,including:the proportion of negative sentiment reviews on micro-blog,the query degree of micro-blog comments by constructing keyword word set,and the proportion of forwarded micro-blog that do not express opinions,and fourteen features are extracted.Random forest,XGBoost,and stacking fusion model are used to identify rumors on micro-blog,and combined the fact,considering the imbalance of the data,and the unbalanced data is tested from the data level and the integrated sampling technology.The final conclusion shows that compared with using only a single classifier,the classification effect of the stacking fusion model has been significantly improved,and the model accuracy rate has reached 95.53%.For the validity of the deep features of micro-blog comments,the accuracy of the model increased by 0.76%after adding micro-blog negative sentiment,and the accuracy rate of the model increased by 2.072%after adding the greed of doubts on micro-blog comment,and the accuracy rate of the model increased by 0.3%after adding the percentage of reposted micro-blog that do not express opinions,it shows that deep feature extraction based on micro-blog comments can effectively improve the rumor recognition effect of the model.