With the increasing use of social media around the world, companies are taking advantage of social media marketing to improve sales. Our idea was to investigate this growing trend using Quandl's datasets and to develop a model for stock trading based on social media engagement data.
What it does
Takes engagement data from Facebook and Twitter for companies in the Movies and Entertainment sub-industry and then uses linear regression to predict the future prices of stock.
Challenges and Accomplishments
It was our first time using Pandas, Scikit, and Git hub for such a project. Coming from a background of engineers and mathematicians, the whole experience at Algothon has helped us learn a lot about working with git, data analysis, and developing a project pipeline.
Our model that we initially built yielded an accuracy of 97%. However, after closer inspection, we determined that the high accuracy of this model was driven by spurious elements more so than any predictive power. We had to redesign our modelling approach to combat this.