Sentiment Analysis Candidates of Indonesian Presiden 2014 with Five Class Attribute

Abstract

Nowadays, Twitter is not only used for social media to maintain friendship, but also Twitter is used to promote and campaign. Twitter usersare free to express their opinions, including opinions about candidates of Indonesian President 2014. This research accommodate the public opinions by classified it into five class attributes : very positive, positive, neutral, negative and very negative. The classification process using Naïve Bayes Classifier (NBC) with data preprocessing using tokenization, cleansing and filtering. The data used in this research are in Indonesian tweets about candidates of Indonesian President 2014, with 900 tweets of dataset and distributed to five class attributes equally. As result, highest accuracy obtained when the experiment using combination of tokenization n-gram, stopword list WEKA and emoticons, which is the values consisting 71,88% accuration, 71,6% precision, 71,9% recall, 6,1% TP rate and 65% TN rate.