Detecting Non-Credible News Using Machine Learning

Many of us are connected to various social media outlets such as Facebook and Twitter. Social media is an online place where one could share interesting information, get updated about current events, and read people's reactions. However, “click-baits”, satirical, false, or misleading articles can result in people making wrong judgements and decisions. The problem is how can people validate and judge information to be credible and maybe non bias before and after they are posted on social media. The project involves using TF-IDF features from articles of credible/non-credible sources to train a model. The model is then validated by evaluating the accuracy of train and test set. The goal is to build a framework that detects non-credible sources and potentially filter them into groups consisting of if the source is trustworthy, or if it appears as though the story itself uses language that would indicate it is to be taken with an open mind. By using news sources and stories that is predetermined to be false, a model can be trained to look for similar language and predict if new articles follow these same patterns. The completion of this framework will lay the groundwork for developing a recommender system that could possibly suggest a balanced list of “left and right” articles based on the information you are reading.