Social Engagement Analysis For Detection Of Fake News On Twitter Using Machine Learning.

Fake news alludes to the news containing deluding or manufactured information that is groundless, which is purposefully misrepresented and spread among a community. It can mutilate reality and become a cause of many social problems and misunderstandings. Many studies have been conducted to investigate the engagement features. However, the choice of parameters is diverse, secondly, the spreading of news is a continuous process that needs to be evaluated. This study proposes a model for analyzing social engagement features on a news dataset of Twitter which is one of the most important platforms for spreading news. The news data was collected from an open-source data store of Kaggle and LIAR. The data was preprocessed and transformed into proportional values with six engagement features i.e. Retweets, Likes, Comments, Quoted Retweets, Multimedia and Images. These features were divided into 05 classification models with a combination of two engagement features. Based on the results four models were selected for further analysis. Models 1 (Retweet and QRT) and Model 5 (Likes and QRT) showed an imbalance of accuracy metrics on fake and real news data. Model 3 (Comments and QRT) and Model 4 (Retweets and Likes) obtained a good balance on accuracy metrics. Results showed that social engagement features can be used to estimate the credibility of news on Twitter. [ABSTRACT FROM AUTHOR]