The Role of Machine Learning in Identifying Fake News and Detecting Anomalies According to a Survey

Abstract

Digital media has contributed to the spread of fake news. The large amount and speed of information spread is creating major societal, political, and security challenges, in which manual methods of fact-checking have become outdated and instead require automated and scalable solutions, that is why, this survey paper explores the role of machine learning (ML) in achieving the goals of dealing with this problem, with two interconnected areas, namely te classification of fake news and the detection of abnormal propagation patterns. We evaluated different models of machine learning as well, i.e., Naive Bayes, Random Forest, K-Neighbors, and Long Short-Term Memory (LSTM) networks on a selected collection of news articles. Our methodology included a three-step process of modeling, along with Natural Language Processing (NLP) steps that included Word Embeddings (Word2Vec) and Word Cloud Term Frequency analysis. Moreover, these findings also revealed that ensemble algorithms, such as the Random Forest, demonstrate a higher level of performance; i.e., their accuracy and F1-score reach 99 percent and 0.99, respectively, which is significantly higher than other algorithms. We also discuss the use of similar ML-based methods to detect anomalies in general and the section about critical issues, such as the bias of the datasets used, the adaptability of the model, and adversarial attacks, and suggest some ways of how such improvements can be reached in the future, including hybrid models and blockchain-inc Because of that, our analysis comes to a conclusion that ML provides a powerful means of improving the integrity of the information, but their effective implementation demands ongoing adjustment and the integration of the multiple disciplines.