FAKE NEWS DETECTION USING MACHINE LEARNING

This project tackles the widespread issue of fake news, which undermines public trust and disrupts informed decision-making. By developing an automated machine learning-based system, it offers a scalable solution to classify news articles as real or fake. The challenge of misinformation, exacerbated by the inefficiency of traditional manual detection methods, necessitates innovative approaches. Using a dataset of 20,800 labeled articles, the system employed a TF-IDF Vectorizer for feature extraction and logistic regression for classification. This combination proved effective, achieving an impressive 95% accuracy, with high precision, recall, and F1 scores, demonstrating its robustness. Minimal false positives and negatives, as confirmed by the confusion matrix, further validated the model's reliability. A user-friendly Streamlit interface enabled real-time predictions with probability scores to ensure practical usability. While the system has shown strong performance, future enhancements will involve incorporating multimodal fusion techniques by integrating additional data sources, such as images and metadata, alongside text. This approach aims to improve precision and generalization, enabling the model to detect fake news more effectively across diverse formats. This work significantly contributes to combating misinformation, providing a reliable and scalable tool for fostering trust in digital information.