Fake News Detection Model Ndetect Using Ensemble Machine Learning Techniques

Abstract: Social media's ability to disseminate information quickly has made it possible for both fake and legitimate news to proliferate, endangering political processes, public confidence, and personal reputations. This paper addresses this issue by putting forth the ensemble machine learning model NDetect, which successfully detects fake news from textual content by combining Decision Tree, Support Vector Machine, Logistic Regression, and Random Forest classifiers. The ISOT Fake News Dataset, which included more than 44,000 news stories with labels, was used to train and test the model. Enhancing the accuracy of fake news identification by utilizing the capabilities of many classifiers through ensemble learning was the main goal of this study. The conventional measures of accuracy, precision, recall, F1-score, and ROC-AUC were used to thoroughly assess NDetect's performance. With a ROC-AUC score of 0.957 and an accuracy of 89%, the suggested model outperformed all baseline models, including Random Forest (52%), SVM (54%), Logistic Regression (53%), and individual Decision Tree (87%).