Machine Learning-Based News Trustworthiness and Type Classifier

With the rise of misinformation online, detecting fake news has become increasingly important. This study presents a smart, reliable system designed to identify fake news and categorize real news into topics such as politics, entertainment, and travel. Our approach combines the strengths of traditional machine learning with those of advanced deep learning models such as LSTM and BERT, helping the system understand the deeper context behind the text. We built a hybrid model using a CNN and BiLSTM with GloVe word embeddings to capture both word order and meaning. To improve accuracy, the system also analyzes related images and audio-since fake news often includes misleading multimedia. It even distinguishes between content written by humans and that generated by AI, addressing modern challenges in misinformation. The system runs on a simple, user-friendly Flask web app that allows real-time detection and classification. It performs strongly, with an F 1 score of 0.98, an accuracy of 95 %, and a loss of just 0.14. In the future, we plan to make the model even better by updating it with new data regularly and adapting to how fake news evolves. Our goal is to offer a powerful, easy-to-use tool for navigating today's complex online information landscape.