Fake news detection using title, URL and tweet and retweet

Abstract

This project investigates the role of social media as a vector for the dissemination of fake news, focusing on the use of URLs, titles, and Tweet recounts to detect misinformation. My research involves developing a machine learning framework capable of identifying potential fake news in social media posts. The main goal of this study is to establish an accurate model that can differentiate between credible and non-credible information based on the features of social media posts, such as the source URL, the structure and phrasing of titles, and the patterns in the spread of the content, as reflected in Tweet recounts. To undertake this task, I pre-processed a dataset of 23,197 social media posts from Kaggle, incorporating their URLs, titles, and dissemination metrics. I utilized a variety of text and data representation techniques to convert these attributes into a format amenable to machine learning analysis. The study harnessed numerous machine learning algorithms, which were refined through hyperparameter tuning, feature engineering, and the use of ensemble methods to boost the predictive accuracy of the model. Through meticulous evaluation and contrasting the performance of multiple models using a range of metrics, the research pinpointed the most effective model for fake news detection. The findings underscore the transformative potential of machine learning and data science in combating the spread of fake news, offering opportunities for early detection and mitigating the impact of misinformation. The ambition of this project is to pave the way for advanced research in the realm of fake news analysis, integrating more complex data sources, and evolving the model for pragmatic application in the dynamic landscape of social media.