Abstract

<p>Fake news, defined as intentionally false information circulated through digital platforms, poses a significant threat to <br>information integrity, particularly on social media. Given its persuasive nature, it is critical to develop efficient methods for <br>detecting fake news and promoting more responsible consumption of online content. This paper proposes a model that <br>leverages supervised machine learning algorithms to detect fake news in online sources. The dataset used in this study, <br>"fake_or_real_news," was preprocessed through feature extraction, narrowing down the data to two columns: text and labels. <br>The label column was further processed to mark true news as "REAL." The study applied three different supervised machine <br>learning algorithms for model training: Logistic Regression, Support Vector Classifier (SVC), and Multinomial Naive Bayes <br>(MultinomialNB). The models were evaluated based on accuracy, with Logistic Regression achieving the highest accuracy at <br>91.9%, followed by MultinomialNB at 88.5%, and SVC at 86.8%. Based on the results, the study recommends using Logistic <br>Regression as the most effective algorithm for detecting fake news in online environments. This approach offers valuable insight <br>into how machine learning can be employed to combat misinformation</p>