This project was created as a part of the class, STA160, to demonstrate statistical analysis, model building, and front to back end pipeline. We used 2 machine learning models to predict whether or not a news article or tweet was true or false and compared our models’ predictions to ChatGPT. The intention is for the user to input text in the appropriate place for a tweet or news article and receive output that shows the model’s truth value prediction as well as ChatGPT’s. From there, the user can make a more educated decision about whether to trust what they’ve read online.
With the current political climate and information arms race between the left and right, the question of truth in online information is often overlooked.
Some may be inclined to believe the average American gets their news from established sources with regulations on what they can claim, but, in fact, one out of five Americans says they regularly get news from news influencers on social media (Stocking, et. al).
Furthermore, 23%--almost a quarter–of people report sharing a made-up news story with and without their knowledge (Barthell, et. al).
Our goal is to build a fully functional machine learning pipeline that helps users
assess the credibility of online content — whether from tweets, blogs, or news articles.
This project classifies text into Real or Fake using two models:
To extend the analysis, we also integrate ChatGPT as an independent predictor. This allows us to compare traditional ML models with modern LLM-based reasoning.
Through our dashboard, users can:
This project was created for STA160 – Fall 2025 at UC Davis, as a demonstration of real-world machine learning workflows, cloud deployment, and web dashboard visualization.