A Machine Learning based Approach for Detection and Classification of Fake News

Abstract

Introduction: Fake news, propelled by the surge in social media and online platforms, poses a significant threat to public opinion, societal stability, and democratic processes. Its historical roots in misinformation, propaganda, and hoaxes have evolved with the virtual age, exploiting the unregulated nature of the internet. Social media's accessibility and rapid information sharing contribute to the swift dissemination of false information, impacting public perception, decision-making, and even elections. The consequences include panic, confusion, and the potential incitement of violence, emphasizing the urgent need for robust detection systems.

Objectives: Detecting and countering fake news on Twitter is essential for information accuracy, utilizing advanced algorithms and machine learning. The primary goal is to categorize news data, differentiating between authentic and fraudulent content, fostering a trustworthy information environment. Validating the legitimacy of news on social media, employing robust verification methods, aims to protect users and promote digital literacy.

Methods: Machine learning techniques, commonly using supervised methods, enhance automated systems for fake news detection by analysing various textual elements like headlines, hashtags, and metadata. Incorporating tokenization and TF-IDF, these algorithms explore term relationships in documents, extracting features to distinguish between accurate and false information. Preprocessing involves noise removal, and feature extraction streamlines the dataset, leading to efficient supervised learning, enabling classifiers to discern between authentic and fraudulent news, providing a valuable tool for media and social platforms to combat misinformation.

Results: Achieving a remarkable accuracy, the model employs the Passive Aggressive algorithm with specific parameters for detecting fake news. Utilizing TF-IDF vectors, the system efficiently classifies news articles as real or fake, demonstrating its effectiveness and proficiency in fake news detection. This high accuracy indicates the model's robustness and capability in accurately classifying and detecting misinformation.

Conclusions: Social media's increasing influence has led many millennials to favor it over traditional news sources. However, this shift has also fueled the rapid spread of false information online, posing risks to journalism and democracy. Misinformation, often driven by sensationalism, can quickly go viral, impacting public perceptions and decisions. To counter this, a machine learning-based approach with a good accuracy rate is proposed. This method efficiently categorizes and detects fake news, safeguarding social media users from deceptive content and its potential societal harm.