Avaliação de Desempenho de Mecanismos de Machine Learning para a Detecção de Fake News

Abstract

Nowadays, a great amount of information is being brought through the internet, especially on social networks, thus providing the user with greater ease of access in obtaining information and knowledge about the most comprehensive subjects. And due to this huge volume of data and information, users Internet service providers become more susceptible to so-called fake news or Fake News (FN), and the harm associated with them. The NF which is defined as the act of spreading false news, rumors, or disinformation through social media, newspapers, television, or radio, has been the cause of major problems, from financial and even issues related to offense, reputation, and dignity. With that even large organizations have already been attacked with the damage that the FN causes. In this context, according to the related literature, several approaches have been developed for the prevention and detection of FN. Among some of the most promising, there is the advent of technologies based on Machine Learning (ML), or from English, ML. With this type of approach, several techniques such as Decision Tree, Naive Bayes, Neural Networks, and Random Forest verify the veracity of certain news that is disclosed and shared on social networks. And yet, several works in the literature have been conducted to mitigate the effects of FN, using ML. In this way, to evaluate these various mechanisms, and to highlight the results that were most satisfactory in the average of the two scenarios the Decision Tree technique, followed by Random Forest, Networks Neural and with lower values than that of Naive Bayes, this work proposal proposes to carry out an empirical evaluation of ML methods applied to NF detection.