Fake News Detection:A Machine Learning Approach

Fake news detection poses a significant challenge in the digital era, characterized by the proliferation of social media platforms and direct messaging systems. This study achieves an accuracy of 98.76% using a Neural Network model and 95.32% with a Naive Bayes classifier for classifying news as real or fake.The neural network, implemented as a feedforward model using TensorFlow and Keras, employs advanced techniques such as dropout layers and regularization to combat overfitting. The text data is preprocessed using tokenization, lemmatization, stop-word removal, and transformed into numerical features using term frequency-inverse document frequency (TF-IDF). This approach demonstrates superior performance, particularly when applied to content-rich datasets, making it a robust and scalable solution for fake news detection.In contrast, the Naive Bayes classifier, while computationally efficient and straightforward, achieves lower accuracy levels compared to the neural network, particularly when dealing with complex or large-scale datasets. Despite this, it remains an effective option for resource-constrained environments due to its simplicity and speed.The results of this study underscore the advantages of neural networks in achieving high accuracy and adaptability, positioning them as a reliable choice for fake news detection in scenarios requiring precision and scalability.