Penerapan Support Vector Machine dengan SMOTE Untuk Klasifikasi Sentimen Pemberitaan Omnibus Law Pada Situs CNNIndonesia.com

Abstract

The declaration of the omnibus law reaped the pros and cons in the community. In a situation like this, the media should be neutral. One of the media that still maintains neutrality is Detik (Rumata 2017). Detik owns several channels such as detikNews, detikFinance, and CNN Indonesia. In this study, the neutrality of the CNN Indonesia media as part of Detik will be studied based on the tendency of sentiment on the omnibus law-related news. Sentiment analysis is used to examine the trend of opinion on news headlines. In conducting sentiment analysis, a method that supports classification is needed. The classification method that will be used in this research is the Support Vector Machine (SVM). There is an imbalance of data in the three categories of sentiment so that the Synthetic Minority Oversampling Technique (SMOTE) method is used to overcome this imbalance. The omnibus law tends to be reported neutrally by CNNIndonesia.com site. The one vs all method has a better classification result than the one vs one method. The application of SMOTE only gives slightly better results than data classification without the application of SMOTE because the imbalance in the data is not too extreme. Modeling using the one vs all method with SMOTE and distribution of data 90% train data 10% test data gives the best classification results with a macro average f1-score of 60,33%.