Applying Ensemble Machine Learning Techniques for Fake News Identification

Abstract

The sharing of information has entered an unprecedented era in human history due to emergence of the Internet With the widespread use of social media sites like Facebook and Twitter. As these platforms enjoy extensive use, users are generating and disseminating a wealth of information, some of which is inaccurate and devoid of factual basis. Detecting false or misleading information within textual content poses a significant challenge. Before arriving at a judgment regarding the accuracy of an article, it is imperative to consider various factors within a specific domain. This paper proposes an Ensemble method for the identification of fraudulent news stories. We leverage different textual features found in both authentic and fake news articles. Our dataset comprises 72,134 news articles, with 35,028 being genuine and 37,106 being false, categorized as binary 0s and 1s. To evaluate our approach, we employed well-known machine learning classifiers including Logistic Regression (LR), Decision Tree, AdaBoost, XGBoost, Random Forest, Extra Trees, SGD, SVM, and Naive Bayes.

To enhance the precision of our findings, we devised a multi-model system for identifying fake news the Ensemble approach and the aforementioned classifiers. Experimental analysis conclusively demonstrates that our suggested ensemble learning technique surpasses the performance of individual learners.