Inspiration:

Fake news has been a growing issue in the past few years. Many companies get targeted by trolls who post fake news which can have damaging impact on their financing.

What it does:

FibStock is an educational attribution platform designed to help users understand the impact of fake news on company stock prices that often leads to investors to be concerned about direction of the business. User can search up a company and see the impact of fake news on said company stock price and related fake news articles that lead to that impact.

Accomplishments that I'm proud of:

It's no surprise this was a challenging task as we had to collect data for fake news from different sources and differentiate the real news from it. We had to change our models to integrate the platform fully. In the end we felt proud as we were able to complete this challenge.

How it works:

We got inspired by open source project jasminevasandani/NLP_Classification_Model_FakeNews to train our model from actual news data. News Title is used as vectors to determine if the news is fake. We used 25000 news records to train our model and ran the model over 1400000 records. Given the tested records we were able to determine a 67~91% accuracy of whether the news is fake or not. Using reddit API and the subreddit news we collect related news about the company and analyze it with our model. Once the results are in, News titles are sent to AWS Sentiment Analysis to find out the nuance of the news and show it as a donut chart. At the same time, We call stock price api to show the price change of the moment that news was published and show it as graph.

Share this project:
×

Updates