Fake News Detection using Machine Learning

By gathering data from a dataset consisting of over 72,000 news articles, this Machine Learning algorithm predicts whether an article is telling the truth or not.

Comment

Inspiration

Information on the Internet is decentralized, world-wide access to read/edit content and allows for the formation of echo-chambers. Fake-News Websites who are more interested in generating clicks for ad-revenue than informing the public (InfoWars, The People’s Voice, Palmer Report) and are harder for general public to distinguish.

What it does

This Machine Learning brain uses a large dataset of news articles with knowing whether they are real or false and uses that to determine whether the news article you are reading currently is telling the truth or not.

How we built it

We collaborated on deep note and worked together to implement the algorithm to determine the article's classification

Challenges we ran into

We had to find an alternative to raw Probability which would be log(Probability) which is more beneficial as the raw probabilities were extremely low and there was a likelihood of them truncating

We also didn't allow the Comparison of Close Probabilities If the probability of the Fake Title and True Title were too close, the determinant would arise from the greater of the probabilities of Fake Content and True Content

Accomplishments that we're proud of

We are proud of our 80% percent accuracy as well as implementing a sort of user-friend way of just inputing a URL and receiving an answer rathe than inputing a file or text document.

What we learned

We learned about how to handle very low probabilities by using logs aswell as how to scrape data from news articles/websites.

What's next for Fake News Detection using Machine Learning

In the future, we could implement this as our own browser or even just a chrome extension. We can have news networks receive a certificate if it passes this test and thus, gain more trust with the public (especially new news networks). Have code for it to constantly increase its database using known-true and reliable news anchors which could be human verified as well.

Built With

Share this project:

Updates