FAKE NEWS DETECTION SYSTEM USING MACHINE LEARNING

Abstract

This project provides a thorough investigation of the identification of false news, including everything from
early data preparation to the creation of several models and a smooth interface with the Flask framework for
user interaction. Importing packages and combining databases of authentic and fraudulent news are the first
steps in the process. The next step is thorough data processing, which includes text cleaning, duplication
removal, and smart Matplotlib and Seaborn display of dataset properties. Feature selection, data splitting, and
tokenization using Tfidf Vectorizer open the door to building a wide range of classification models, from SVC
and Logistic Regression to KNN and Random Forest and cutting-edge ensemble techniques like Voting
Classifiers and Stacking. User registration and signin features are facilitated using the Flask framework in
conjunction with SQLite.
Results are shown on the front end after user-inputted text is translated and preprocessed for prediction using a
trained model. Moreover, new models are shown, such as the Voting Classifier (XGB+PA+Boosting), which
exhibits an astounding 100% accuracy. Notably, the initiative improves the user experience overall by allowing
users to submit text for real-time predictions on the frontend. The project is positioned as a significant addition
to the changing field of information integrity due to its comprehensive approach to false news identification,
which includes user-centric design and methodological variety.