Tri Rivanie, Rangga Pebrianto, Taopik Hidayat, Achmad Bayhaqy, Windu Gata, Hafifah Bella Novitasari


The pandemic that occurred in Indonesia has not yet subsided and far from under control. Indonesian Ministry of Health is most appropriate person to responsible for providing an explanation of actual situation and extent to which state has handled it. However, he has rarely appeared in public lately to explain about handling of Covid-19 pandemic. In response, many people are pros and cons come to give their opinions and feedback. The increasing use of internet during pandemic, especially on social media, where one of them is Twitter, which is a means of expressing opinions. Posting tweets is a community habit to assess or respond to events, as well as represent public's response to an event, especially Ministry of Health steps and policies in handling and breaking chain of Covid-19 pandemic.

The tweet posts were taken only in Indonesian-language and also related to performance of Government, especially Ministry of Health. After that, a label is given so that sentiment of tweets is known. To test results of these sentiments, an algorithm is used by comparing two methods of Support Vector Machine (SVM) and Naïve Bayes (NB). Validation was carried out using k-Fold Cross Validation to obtain an accuracy value. The results show that accuracy value for NB algorithm is 66.45% and SVM algorithm has a greater accuracy value of 72.57%. So it can be seen that SVM algorithm managed to get the best accuracy value in classifying positive comments and negative comments related to sentiment analysis towards Ministry of Health.


KeywordsSupport Vector Machine, Naïve Bayes, Analisis sentimen, K-Fold Cross Validation

Full Text:







Jl. Zainal Abidin Pagar Alam No. 93 Labuhan Ratu, Bandar Lampung.Kampus IBI Darmajaya,  Gedung A Lantai 2.Telp. 0721-787214, 781310 Fax. 0721-700261 ext.126

Index by:


Flag Counter

Creative Commons License

Jurnal Informatika is licensed under a Creative Commons Attribution 4.0 International LicenseMy Stats jurnal