Research And Implementation Of Fake News Detection Based On Deep Learning

At present,the production sources of news information continue to increase the volume of fake news is also increasing and its negative impact on society is growing.The fake news involved in this study is mainly news of the rumor type,and fake news usually have differentiated emotional characteristics,which are used to strengthen the dissemination and influence of information among users.Based on this,many studies model the emotional characteristics of news content,analyze the intention of news expression,and then judge the authenticity of news.While these studies have certain postivie achievements,there are also several problems:Firstly,these studies are basically limited to the analysis of news content while ignored the auxiliary role of news comments in news detection research as the user's intuitive emotional feedback on news content;Secondly,the existing data sets are insufficient to support pre-analysis and posttraining,which affects the model ability;Thirdly,it has not yet been able to provide users with an aggregated portal to view and detect news at the same time,and more research is still in academics.Accordingly,this paper will carry out research work from three aspects: This paper investigates the above-mentioned issues in three aspects:1)Regarding the problem of the scope of analysis,this paper will concurrently include news contents and news comments into the test and carry out relevant research.Firstly,regarding news contents,this paper will introduce a self-attention mechanism and combined with LSTM long and short memory units to establish a news context analysis model;secondly,regarding news comments,this paper will optimize the CNN network to establish a comment analysis model.2)Regarding the lack of training samples for news texts and the low robustness of the detection model,this paper introduces an adversarial neural network to optimize and train the news text analysis model.Then,based on the random forest,the paper will fuse the news context analysis model and the comment analysis model to obtain the final detection model.3)Regarding the lack of aggregation entry,this paper,based on the aforementioned final detection model and combined with the demand analysis of the current news system,will establish mobile and administrative fake news detection system based on Java,Python,and Vue.This system implements news data crawling,data cleaning,detection and filtering,cluster display,fake news analysis,and news browsing and other functions.Finally passed the test,this paper completed the system function and performance reliability verification.In summary,the research process of this paper is to apply deep learning to fake news detection and establish a system for landing,which has certain significance for the design of news detection system.