Enhancement of Rumors and Fake News Predictions Based on Machine Learning Techniques

Abstract

Rumor's propagation is accelerating along with the rapid rise in popularity of social networks. Any unverefied statements that originate from one or more sources and subsequently spread throughout the meta-networks are referred to be rumors. Because rumors could occasionally be harmful, especially when it comes to social and political concerns that have a greater influence on people's lives, it is imperative to find strategies for detecting rumors as soon as feasible. This study is aimed at developing a classifier for Rumor Detection using Machine Learning (ML) algorithms and Natural Language Processing (NLP) tools. Four preprocessing types are utilized by the system. The most widely utilized feature extraction (FE) approach is the Term Frequency-Inverse Document Frequency (TF-IDF) technique, which comes next. Rumors are detected using Stochastic Gradient Descent (SGD) approach and five classification algorithms: Naive-Bayes (NB), Random Forest (RF), K-Nearest Neighbor (K-NN), Logistic Regression (LR), and Decision Tree (DT). The best accurate method for classifying rumor data was determined by comparing these six methods. Additionally, recall, precision, F1-score, and accuracy measurements have been used to evaluate the classification algorithms' results based on performance metrics. Results indicated that RF applied to three datasets produced the maximum accuracy, with 93% accuracy in the SNOPES dataset and 93% accuracy in the unbalanced dataset. The SGD approach yielded 96% accuracy for the balanced dataset Faknews

Introduction: Nowadays, social networks are with no doubt the most important for accessing information. Due to the large number of social network relations and the vast amount of data, scientists are working to overcome a number of obstacles resulting from the rich relations that social networks contain [Alatas and Altay, 2018]. Although social networks are an excellent to obtain news, they have become a powerful tool for manipulating people and a source of rumors on any topic. Rumor is a social problem with numerous negative reflections. Positive relations could deteriorate and the company's and individuals' reputations could be harmed. Serious financial harm could be done to individuals, businesses, families, and even entire nations. People have lost a lot of as a result of gossip events, which have become a major issue in social networks recently. For instance, rumors have a detrimental effect on the political landscape, the economy, and the stability of society. A rumor about "shootouts and kidnappings through drug gangs near schools in Veracruz" surfaced on Twitter and Facebook on August 25, 2015. There have been 26 car accidents in the city as a result of people leaving their cars in the middle of the road and hurrying to pick up their kids from school [Tang]. The necessity of the automatic prediction of authenticity of content on social media has been shown by this case of false rumors. A lot of research into creating systems that can identify rumors on their own depends on AI tools like NLP and ML..

Objectives: This paper aims to design and implement a Rumor Detection System with the use of ML and NLP. The system is designed for accurately classifying posts on Facebook and Twitter as either truths or rumors. It uses different ML algorithms, such as RF, NB, LR, K-NN, DT, and the SGD. The aim is achieving high accuracy, aiding opinion analysts in detecting and classifying rumors across diverse datasets.

Methods: At varius vel pharetra vel turpis nunc eget lorem. Feugiat scelerisque varius morbi enim nunc. Cras semper auctor neque vitae tempus quam pellentesque nec. Faucibus purus in massa tempor nec feugiat nisl. Congue nisi vitae suscipit tellus mauris a. Est sit amet facilisis magna etiam tempor. Dictum varius duis at consectetur. Purus semper eget duis at tellus at urna. Ipsum consequat nisl vel pretium. Viverra maecenas accumsan lacus vel facilisis volutpat est. Bibendum arcu vitae elementum curabitur vitae nunc sed. Nisl tincidunt eget nullam non nisi est. Ac turpis egestas integer eget aliquet nibh praesent.

Results: Egestas diam in arcu cursus euismod quis viverra nibh. Convallis aenean et tortor at risus viverra. Sit amet justo donec enim diam. Sem et tortor consequat id. Purus gravida quis blandit turpis. Consectetur adipiscing elit duis tristique sollicitudin nibh sit amet commodo. Eget duis at tellus at urna condimentum mattis pellentesque. Auctor elit sed vulputate mi sit amet. Consequat ac felis donec et. In dictum non consectetur a erat nam at lectus. Dui vivamus arcu felis bibendum ut tristique. Lacinia quis vel eros donec ac. Ac turpis egestas maecenas pharetra convallis posuere morbi leo. Tortor id aliquet lectus proin.

Conclusions: Mi tempus imperdiet nulla malesuada. Magna fermentum iaculis eu non diam phasellus vestibulum. Consectetur adipiscing elit duis tristique sollicitudin nibh sit amet commodo. Elit scelerisque mauris pellentesque pulvinar. Et malesuada fames ac turpis egestas maecenas pharetra convallis posuere. Elementum integer enim neque volutpat ac tincidunt vitae semper.