A Fake News Classification and Identification Model Based on Machine Learning Approach

Abstract

In the recent past the popularity of the social media platform has increased exponentially and at the same time various challenges have also been increased. One of the major challenges is related to fake news on social media platforms. It is really nontrivial task to filter and distinguish between fake and the real news. In this paper, various machine learning models have been applied to identify and examine the fake news on social media platforms. The Naive Bayes, Support Vector Machines, Passive Aggressive Classifier, Random Forest, BERT, LSTM, and Logistic Regression, were used to classify and identify the fake news on various social media platforms. The work is based on an ISOT dataset of 44,898 news samples gathered from a variety of sources and pre-processed with TF-IDF and count vectorizer. On evaluating the performance of algorithms on the given dataset, it shows that the precision of the Passive Aggressive Classifiers is 99.73%, Naive Bayes is 96.75%, Logistic Regression is 98.82%, BERT is 97.62%, LSTM is 97.44%, SVM is 99.88%, and Random Forest is 99.82%. Therefore, it is concluded that the SVM is one of the best performing algorithms in terms of precision to identify the fake news on social media. However, there are very marginal differences in the performance of the SVM, Random Forest, and Progressive Aggressive Classifiers in terms of precision. Further, an algorithm can be designed and developed to collect the news available on the various social media platforms to maintain the dataset in real time and analyze the same to identify the fake news.