EVALUATION OF THE PERFORMANCE OF SVM AND NAIVE BAYES METHODS IMPLEMENTED WITH EMOTION CLASSIFICATION FOR SENTIMENT ANALYSIS ON TWITTER DATA

RANI RIZKIANI ILYAS, Khabib Mustofa, S.Si., M.Kom., Dr. techn.

2020 | Skripsi | S1 ILMU KOMPUTER

Sentiment Analysis is one of the most used methods in order to gather emotions through microblogging websites, particularly Twitter. However, the usage of Emotion Classification for Sentiment Analysis, particularly in the Indonesian language, is still rare, particularly with the usage of SVM and Multinomial Naive Bayes.

On popular microblogging websites such as Twitter where users are free to express their emotions, the concept of emotion as well as its classification is very apparent on texts. While the concept and study of emotion classification is quite crucial for the analysis of various topics in order to improve sales and general public opinion, research regarding the use of emotion classification for e-commerce websites is exceedingly rare. Moreover, research surrounding the development and analysis of emotion classification for tweets in other languages, particularly Indonesian, is still quite lacking. This research attempts to study the use of emotion classification that is available on Indonesian tweets, as previous researches has only used Support Vector Machine. this research attempts to use other methods of classifying the text such as Multinomial Naive Bayes as well as Support Vector Machine to see which classification method is best used for emotion classification. This research also aims to increase the precision that was established by creating a new emotion classification dataset that is only specified around e-commerce related tweets. The dataset consists of a dataset with e-commerce related tweets from three Indonesian e-commerce applications such as Shopee, Tokopedia and Bukalapak. The extraction feature method that is used is TF-IDF and the performance measure that will be seen is the precision score. With the highest accuracy score of 70% for Support Vector Machine, and 65% for the highest score for the Multinomial Naive Bayes.

Kata Kunci : Support Vector Machine, Multinomial Naive Bayes, Emotion classification, Text-Classification, TF-IDF