There are few things in the world like dedicated college basketball fans: their knowledge and passion for their teams are on full display every March as dozens of teams battle for the championship. When you add in the thrill of predicting the games through bracket competitions, it all adds up to an intense few weeks of competition.
We decided to join in on the fun by using our Smart AI™-powered Experience Analytics Platform to help us predict the outcomes of the games. With so many models at hand to work with in the platform, we took several approaches to try to predict who would win each game.
In today’s article, we’re going to focus on sentiment analysis of social media posts related to each of the 16 final teams. By analyzing Twitter posts in the days leading up to the game, we wondered whether we could predict the outcome by measuring the sentiment around each team on social media.
Sentiment analytics use Natural Language Processing to score the emotional register of texts. While humans are great at understanding the emotion behind language, advances in artificial intelligence makes it possible for machines to analyze large volumes of data quicker than any human. It can be applied to any form of customer feedback, from social media posts and call center transcripts to chats and emails (Read more about sentiment analysis here.)
Our hypothesis is that the more positive the overall conversation was surrounding each team, the more likely it was that fans’ thought that team would win. We thought of it as a way to crowdsource the “eye test” or the “gut feeling”— sort of like how I “knew” Oral Roberts University would pull off its first round upset by looking at their season performance. (Of course, I failed to predict the rest of their run in my own bracket.)
The short answer is, no, fan sentiment can’t help predicting outcomes. The long answer is also no. Even though fans have a finger on the pulse of their favorite teams and our sentiment analysis models correlated 5 out of 8 teams correctly in the round of 16 (63%), this has been the most unpredictable tournament in history.
Since we were relying on fans’ emotions and intuition, the data just isn’t enough to predict future behaviors of the teams themselves. But there is data that can help us predict likely outcomes, and that’ll be the focus of our next post. In it, we’ll show you how we used our Auto Learn models with historical team performances to analyze the head-to-head battles.
Luckily, our Smart AI™-powered sentiment analysis is great at many things beyond predicting sports outcomes. Sentiment analysis is a terrific way to uncover consumer perceptions and opinions. By correlating emotion with trending topics automatically, you can help cover any blind spots you might have missed from surveys and other forms of feedback.
With so much data out there to get insights from, having the help of artificial intelligence to make sense of it all can free up your teams to focus on acting on insights instead.
What else can sentiment analysis do? It can help you protect your brand on social media by tracking sentiment towards products and automatically flagging worrisome trends. It can help you mine for competitive intelligence by seeing what consumers are saying about the competition. And it can help you uncover emerging trends and emotions that might point to issues that can help you predict likelihood for customers to churn.
Check out below to see the results of each head-to-head analysis.