Twitter Sentiment Analysis using Na¨ive Bayes Classifier with Mutual Information Feature Selection

Abstract

Sentiment analysis is an identification technique of emotion expressed in texts. The
sentiment analysis goal is to determine a negative or positive opinion within a sentence
or a document. Twitter is one of social medias to convey an opinion. The twitter allows
its users to write opinions related to a specific topic in a tweet. The twitter data used in
this research was downloaded using the twitter Application Programming Interface (API).
It consisted 500 tweets about Lombok tourism that contained #lombok and
#woderfullombok hashtags. The features extracted from the twitter data were selected
using the Mutual Information (MI) method then they were analyzed using the Naïve
Bayes Classifier (NBC) model. The evaluation of sentiment analysis on the Lombok
tourism twitter data in a 10-folds cross validation resulted 97.9% accuracy.
Key words : Sentiment Analysis, Twitter, Naïve Bayes Classifier, Mutual Information.