Fake News Detection Using Logistic Regression, Sentiment Analysis and Web Scraping

Abstract

    In recent years information sharing through internet and social media have grown abundantly and it is hard to find the authenticity of the news.  This paper aims to detect the fake news using machine learning and web scraping methodologies. The proposed methodology performs the fake news classification on the labeled dataset based on one of the Machine learning algorithms i.e. logistic regression. The proposed system also scrapes multiple news websites and checks for keywords mentioned in the input news article and calculate a percentage combining the results from the machine learning approach which uses the logistic regression algorithm and the web scraping approach. Also, a major factor to weigh in terms of fake news is the sentiment or the emotion behind the news article, the news article may be biased concerning a specific political propaganda, hence in favor of a side. But true news must be unbiased; it must be factual, not accusing or declarative. Thus, to inform the user about the bias in the news article, the proposed system performs sentimental analysis involving pre-processing, NLTK lemmatization, porter stemmer and uses naïve Bayesian algorithm and then presents the user with the sentiment behind the news article. The two above mentioned process will present the user with two outputs: the percentage of how accurate the news article combining the results from the logistic regression algorithm and web scraping approach and will output the sentiment behind the content through sentiment analysis. This will facilitate the user to take an informed decision about the nature of the news article given by the user.