Civil Aviation Events Public Opinion Trend Forecast Method Research Based On The Analysis Of The Text Orientation

People has paid more and more attention to the civil aviation of China with its rapid development,and public events of civil aviation are highly concerned due to some new media,such as micro-blog,forums.Netizens comment on events of civil aviation by these platforms,but some comments have nothing to do with the topic and some are unreal.So it is important to filter spam comments before analyzing events of civil aviation.In addition,it is also significant to analyze comments emotions and forecast trends about comments which affect the attitude of others.Six indicators are defined for recognizing and filtering spam comments which include whether exists repeated comments and the number of government department's occurrences,etc.The information gain algorithm is used to calculate weight of characteristics and the particle swarm optimization support vector machine model(PSO-SVM)is used to identify and filter spam comments.And because the acquisition of predictive indicators are the premise of the network public opinion emotional trend prediction,a new forecast indicator is proposed which is defined comments emotional tendency values time series and it is different from the previous simple heat index(such as attention,comments reply number and forwarding number,etc).Due to emotional tendencies are characterized by non-linearity and randomness,the relevance vector machine model is used to forecast the trend to improve the accuracy.Experiments are designed in this paper,through which analyzing and verifying research results.To the problem of recognizing and filtering spam comments,experiments analyze the effect of characteristic's number of defining spam comments and different characteristics on identifying spam comments,results of experiments describe the importance of choosing the right features for identifying spam review.Comparative experiments which include relevance vector machine,Elman neural network and BP neural network to the emotional trend prediction are carried out.And mean absolute error(MAE)and root mean square error(RMSE)are used to evaluate the accuracy of prediction.Comparative experiments explainthat the predictive performance of relevance vector machine can more accurately reflect the emotional trend of netizens on public opinion events than other models.Therefore,it is meaningful to study the identification of spam comments and the prediction of emotional trend in civil aviation public opinion analysis.