Published Paper Details:

The widespread propagation of fake news via digital means gravely impairs constructive civil discourse, thus making detection systems essential. This work presents a machine learning-based fake news detection framework implementing NLP and text classification techniques. A labeled dataset of real and purposely fabricated news articles is preprocessed by lowercasing, removal of punctuation, and normalization of tokens, and feature extraction is performed using a TF-IDF vectorizer. The dataset is then split into training and testing sets to achieve the strong evaluation purposes. Four classification models of Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), and Gradient Boosting (GB) are trained and tested under the evaluation criteria of accuracy, precision, recall, and F1-score. The experimental results reveal that a pair of ensemble classifiers, namely RF and GB, are the best performers in fake news detection. Further, the system permits manual testing of any custom text inputs-from-the-fly to make it more usable in real-life settings. From the results, it can be inferred that NLP combined with ensemble machine learning can provide a scalable and automated tool for the efficient identification of fake news. Future enhancements can include deep learning and large language models to improve accuracy and the ability to adapt.