Research And Implementation Of Fake News Detection System

With the advancement of Internet technology,the way of people get news has also changed.People are more inclined to obtain news information through network platforms.Compared with traditional news,online news has a faster update and dissemination speed.In addition,the platforms for publishing news are more diverse.While bringing convenience to people,it also makes the authenticity of news has become more uncertain.Fake news has a strong influence.If people mistrust and spread the fake news,it will cause misunderstandings among the people and cause negative emotions.In the worst case,it will affect social stability.Therefore,this article has developed a network fake news detection system to assist news organizations and media platforms to identify fake news in a timely manner,prevent the spread of fake news,and reduce its influence,so as to achieve the purpose of maintaining social stability and promoting the healthy development of social networks.This article carefully analyzes the current status of research on fake news detection at home and abroad,and finds that in feature selection,most studies have extracted some relatively simple statistical-based features,while ignoring the features of the news text itself.Therefore,this article will detect fake news by introducing the theme-based news text sentiment feature,text similarity feature,and the sentiment category features of the comments based on the original features.Then integrate the extracted features as an input feature of the SVM model to train a classifier to realize the recognition of fake news.The main functional modules implemented by this system are the acquisition of news data and comment data,the retrieval and analysis of news data,the fake detection of network news,and the visualization of fake news data.First,the topic words of the news are extracted by the TextRank algorithm,and corresponding fake news articles are found according to the topic words.Secondly,use the cosine similarity algorithm to calculate the similarity between the news text to be detected and the corresponding fake news text,so as to extract the similarity feature of the news text.Then,it is necessary to perform sentiment analysis on the news text and comments to extract the sentiment characteristics of the news text and the sentiment category features of the comments.Finally,the extracted news text features,comment features,and statistics-based user features are constructed into feature vectors.It is also necessary to optimize the c and gamma parameters in the SVM classifier,and use the optimized values to train the data to obtain a detection model.This model is used in the system to detect news data in order to realize the identification of fake news.