Term Frequency Tokenization for Fake News Detection

In today's world, when the internet is pervasive, everyone gets news from a variety of online sources. As the use of social media platforms has grown, news has travelled quickly among thousands of people in a very less duration. The propagation has been far reaching for the fake news generation in repercussions, from altering election outcomes in support of specific politicians, creating prejudiced viewpoints. Furthermore, spammers use appealing news headlines to make cash through click-bait adverts. In today’s world knowingly or unknowingly fake news spreads around the world through internet. This has a great impact on the people who blindly believe whatever the internet provides. Hence, fake news identification has become a new study subject that is attracting a lot of attention. However, due to a lack of resources, such as datasets and processing and analysis procedures, it encounters several difficulties. This research uses a non-probabilistic machine learning models of computational prototypes to address this problem. Furthermore, the comparison of Term Frequency-Inverse Document Frequency (TF-IDF) is done, for the purpose of determining the best vectorizer used for detecting fake news. In order to raise the accuracy, stop words of English are used. To predict bogus news, a Support Vector Machine (SVM) classifier is deployed. According to the simulation data, the SVM and the TF-IDF produce results with high accuracy.