Twitter Sentiment Analysis using Machine Learning

Abstract

As the number of social media users is being higher, many people are sharing various opinions and
each country’s real-time situation by online. Also, the influence of online information is increasing
to such an extent that the individual’s actual behavior or situation can be estimated. In this situation,
researches to analyze through social media are being actively carried out in order to identify problems
in real life. In this research, we proved that we can infer actual behavior or situation based on
individual social media activities. This research focused on the Twitter platform that is actively used
to express individual emotions in social media platforms. We analyzed tweets of Donald Trump and
Hillary Clinton who were the 45th presidential candidates of the United States of America. Several
methodologies like sentiment analysis, topic modeling, and machine learning were used to prove
correlation between Donald Trump’s tweets and his behavior. Through experiment, it proved not
only we can adjust classification and clustering algorithms but also Decision Tree was the most accurate
algorithm. Finally, we proposed the possibility of applying the above method to a system for
detecting anomaly symptoms by concentrating on negative messages. It is expected to provide social
media users with sufficient awareness of online activities.