Fake News Detection on Social Media Using a Stacking Model

Fake news has existed since long before the Internet was invented. It can be defined as “fictitious articles deliberately fabricated to deceive readers”. While some news may have a grain of truth, it frequently lacks any contextual information. They may not contain any sources which can be verified. Some of the reports may contain straightforward factual statements, but they are intentionally worded in an inflammatory sense, often omitting crucial information, or just give one point of view. We tried to understand the nuances in fake news in this research paper by its language characteristics. Initially the news articles content was pre-processed using techniques like Lemmatization and TF-IDF. The data was then classified using a Stacking Classifier, using Naïve Bayes, SVM, and Decision Tree is used as the base models and Logistic Regression is used as Metamodel.