An Exploration of Using Twitter Data to Predict the Results of the U.S. Primary Elections

The use of social media user feeds is a common interest of researchers exploring public views and opinions. In this exploratory study, we look to investigate how Twitter feeds during a presidential primary election can be evaluated to determine the relationships between contesting candidates and garner any predictive insight into election contest outcomes. In this study we collect data from both the REST API and STREAMING API from Twitter, each having their own data collection merits, and perform an association analysis, sentiment analysis, and linear regression to determine what insights can be captured from the data. In this work we find revealing relationships between candidate users accounts on how they interact with each other. We also show how sentiment from verified user accounts on Twitter show significance in election contest outcomes.