A Novel Approach to Ambiguous Fake News Classification through Machine Learning

Abstract

The rise of digitalization and technology has substantially expanded the number of people with online news access, resulting in a significant shift in how information is consumed on social channels. The spread of fake news with ambiguity is a critical and harmful aspect of handling online news access. Such news should be identified quickly since it can harm an individual or organization's reputation and holds the capacity to influence one's actions which can be a potential threat to modern civilization. The proposed work in this paper presents an ambiguous fake news classification model using the decision tree, random forest, and SVM. It first pre-processed the known dataset, extracted its features, and then provided the training to all three classifiers. Further, the classifiers were tested against unknown datasets. The experiments were performed on the collected dataset of 40,000 records including fake and real news. It is observed that it achieved very promising experimental results of precision, recall, and accuracy. It obtained the best results with the decision tree, that is, 0.9977 for both precision and recall along with 99.67% accuracy.