Fake News Detection System
Mr. H. M. Gaikwad1, Mr.S.V.Waghmare2, Sakshi Patil3, Heetal Sarode4, Grishma Sawant5,Sakshi Jadhav6
Head of AIML Dept, K.K. Wagh Polytechnic, Nashik1
Lecturer in AIML Dept, K.K. Wagh Polytechnic, Nashik2
Third Year Students of Artificial Intelligence and Machine Learning, K.K. Wagh Polytechnic, Nashik3-5
Abstract: The Internet has revolutionized the way we communicate and share information. Millions of people use various social media platforms to post and spread news. However, since these platforms do not always verify the identity of users or the authenticity of their posts, they have become fertile ground for the spread of fake news. Fake news can serve as propaganda against individuals, societies, organizations, or political parties, making it increasingly difficult for humans to manually detect misinformation.
Given the volume and speed at which information spreads online, there is a growing need for machine learning classifiers that can automatically detect fake news. The proliferation of social media has profound consequences on society, culture, and business. It both enhances connectivity and, at the same time, facilitates the rapid dissemination of misinformation for commercial or political gain.
Technologies like Artificial Intelligence (AI) and Natural Language Processing (NLP) offer significant opportunities to develop systems that can automatically identify fake news. However, this task is challenging because it requires models that can understand and summarize news content, then compare it against verified information for accurate classification.
This project proposes a robust framework that utilizes deep learning models and advanced NLP techniques. In the context of our implementation, we leverage traditional methods (like TF-IDF for text vectorization) along with a machine learning classifier (such as the Passive Aggressive Classifier) to distinguish between real and fake news. Although the high-level project description mentions the use of sophisticated models—such as improved Recurrent Neural Networks (RNNs), Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and even attention mechanisms with pre-trained models like BERT or GPT—the provided code demonstrates a more streamlined approach focused on effective feature extraction and classification.
