NBA All-Star prediction using twitter sentiment analysis

Abstract

As Web 2.0 services become more popular, social network analysis related research have received more attention. Typically, most Internet users contact others through a variety of social media, such as Facebook or Twitter. This research explores human behavior by conducting opinion mining on Twitter to predict the final voting results of NBA All-Star 2013. The termfeature model is proposed to filter out noise for enhancing the quality of the tweet corpus. Tweenator, an emotion detector, assists to decide whether the emotion tag for each gathered article is positive or negative. Two factors are counted in this research: the number of tweets and the ratio of positive tweets for each candidate player. According to experimental result, the positive tweets has direct ratio with the number of votes in the NBA All-Star Game, a result suggesting that sentiment analysis is an effective tool for predicting human voting outcomes.