A Machine Learning Model to Identify Fake Data from Social Media using Sentiment Features

Abstract

The exponential growth in the use of social media is leading to sharing of information among each other through which the spreading of fake news is common these days. online social networking is the main source for fake news. The most popular social media are Twitter and Facebook, through which the majority of the news reaches the public. This study is aim to try different classification algorithms in comparing with Dataset. For our experiment purpose the dataset used is Real or Fake News dataset which is extracted from Kaggle, which comprises 30Mb of twitter data. The two major classification algorithms used are Naive Bayes and Logistic Regression classification algorithm. The algorithms result in Accuracy score 82.48%, AUC 1.0 and kappa score 0.64 and Accuracy score 91.16%, AUC 0.91 and kappa score 0.82 respectively for the given dataset. The two different classification algorithms are successfully checked with the given dataset. The sentimental analysis is the other way of identification of fake data problem which can be implemented to know the positive and negative sentiment in the given twits. VADER feature is the one of the feature extraction which can be tried with the dataset to find out fake and real data.