Classification of Emotions on Social Media Using Support Vector Machine and N-Gram

Abstract

The large number of Facebook social media users in Indonesia provides an opportunity for researchers to study various things through text, one of which is the translation of emotions through status update text which is the main feature of the social media. The status update feature is often used by the public to share information, express opinions on something or express opinions on experiences that are a trend of public attention. In addition, the status update feature is also often used by the public to share the emotions they are currently feeling. To analyze data from Facebook status updates which are irregular in nature, it is necessary to do text analysis using the text mining method.

The author proposes to conduct research on Facebook status updates using the text mining method. Where in this study using N-Gram level feature extraction and the classification method used is the Support Vector Machine (SVM) method. From the tests that have been carried out using this method it produces an accuracy value of 53%. The highest precision values ??are generated by several emotions, namely fear, shock, disgust, and anger with a value of 100%. Furthermore, the highest recall value is generated by the Happy emotion with a value of 100%. And finally, the highest f1-score is generated by the emotion of shock with a value of 67%.