33Fake news detection

This project aims to detect fake news using both machine learning and deep learning models. It uses a dataset of labeled fake and true news articles, preprocessed by removing irrelevant content such as URLs, digits, and special characters. The text is vectorized using TF-IDF to convert it into numerical features for model training. Several machine learning algorithms, including logistic regression, decision tree, gradient boosting, and random forest, will be trained and evaluated against Precision, Recall, and F1-Score. To capture complex patterns in text, an LSTM neural network is also applied. SMOTE (Synthetic Minority Over-sampling Technique) was used to handle class imbalance and improve the performance of the model. Additionally, a BERT-based transformer model was tested to gauge the potential of advanced natural language processing techniques for detection of fake news. Manual testing allows users to input real-time news articles for predictions from the trained models. The performance results are compared across models, visualized using bar plots, and evaluated using adversarial examples to ensure robustness. This project demonstrates the effectiveness of various machine learning and deep learning methods in identifying fake news, contributing to the fight against misinformation.