Classificação de fake news por mineração de texto

Abstract

False news (known as fakes news) is being intensified in social media and online forums, with the fundamental role of spreading misinformation, influencing, and distorting reality. Disinformation is increasingly present in a society due to the rapid transmission of false allegations on social networks, lack of fact-finding and the population's increased access to social media. In addition, fake news has drawn a lot of attention in the world due to its negative impacts on politics, economy, or even personal ones. Therefore, this study aimed to classify news as true or fake news through text mining techniques. For classification, the data mining tool WEKA (Waikato Environment for Knowledge Analysis) was used to apply the decision tree technique and the resources Bag-of-Words (BOW) and Term Frequency-Inverse Document Frequency (TF-IDF).The experiment was evaluated by accuracy, precision, sensitivity (or recall) and f-measure metrics, and the results sought to be promising. The models were validated with the Fake.br Corpus set, which has 7.200 websites texts that are already pre-processed in Portuguese. The result of this experiment showed that the TF-IDF model obtained the best performance in relation to the BOW, with an accuracy of 89.82%.