DefinePK

Abstract

Effortless access to the bulk of information and its spread is becoming a major issue these days. Many people rely on social networking sites and electronic media to get information. These sites might be used to spread information more quickly, especially fake information. This widespread has catastrophic results on individual and society. Identifying the fake news over numerous mass media platforms have rendered traditional machine learning algorithms less effective. Since fake news detection is vital, this study aims at analysing common machine learning algorithms- linear regressor, decision trees, random forest and Naïve Bayes, MLPC and LSTM and the ensemble methods XGBoost and CatBoost, particularly to validate the efficiency of Kaggle dataset on fake news (fake-and-real-news-dataset). The results reveal a surprising lead of ensemble methods and LSTM over other algorithms in this dataset, particularly the XG Boost Classifier achieved the highest accuracy of 92% in validation and 100% in training. Similarly, Recall score of CatBoost is higher than others.