Misinformation Stocks

Abstract

The growing levels of fake news in our media contributes to misinformation campaigns, the impact of which can spread to investment decisions. To analyze the extent of misinformation on investors, we collected financial articles surrounding specific stocks. We compiled these into a dataset containing the titles, dates, sentiment analysis and misinformation rating. We leveraged a machine learning framework to automatically determine the sentiment of a given article. A value is manually assigned to each article in reference to its level of misinformation. This information is displayed in a digital library for users to access. From our small case study analysis, our preliminary findings show that the misleading content of an article ultimately has little impact on stock value. Instead, the sentiment of the public towards the news, regardless of its validity, is the driving force behind price fluctuation. We created a zip file that includes all of the code that we created for each part of the project, so that others can access and add to what we made. We demonstrated the different aspects of our project in a presentation, explaining briefly how each of the sections work, with pictures to aid understanding.