ANALISIS SENTIMEN TERHADAP VAKSIN NUSANTARA PADA MEDIA SOSIAL YOUTUBE MENGGUNAKAN METODE NAÏVE BAYES DAN SELEKSI FITUR PARTICLE SWARM OPTIMIZATION

• मुख्य लेखक: Taufik Adi Prasetyo, (लेखक)
• स्वरूप: पुस्तक
• प्रकाशित: 2022-07-12.
• विषय: Thesis; NonPeerReviewed
• ऑनलाइन पहुंच: Link Metadata
• Notes: http://repository.upnvj.ac.id/19802/1/ABSTRAK.pdf; http://repository.upnvj.ac.id/19802/2/AWAL.pdf; http://repository.upnvj.ac.id/19802/3/BAB%201.pdf; http://repository.upnvj.ac.id/19802/4/BAB%202.pdf; http://repository.upnvj.ac.id/19802/5/BAB%203.pdf; http://repository.upnvj.ac.id/19802/6/BAB%204.pdf; http://repository.upnvj.ac.id/19802/7/BAB%205.pdf; http://repository.upnvj.ac.id/19802/8/DAFTAR%20PUSTAKA.pdf; http://repository.upnvj.ac.id/19802/9/RIWAYAT%20HIDUP.pdf; http://repository.upnvj.ac.id/19802/10/LAMPIRAN.pdf; http://repository.upnvj.ac.id/19802/11/HASIL%20PLAGIARISME.pdf; http://repository.upnvj.ac.id/19802/12/ARTIKEL%20KI.pdf
• Keywords: QA75 Electronic computers. Computer science

Youtube is one of the many social media that can be used to provide comments that can be accessed by various groups of people, including the people of Indonesia, comments written can be in the form of complaints, suggestions, or constructive criticism on a topic, one of which is about the Nusantara Vaccine. This research uses a video on YouTube entitled "Peneliti Utama Jawab Kontroversi Vaksin Nusantara - ROSI (1)" and uploaded by the KOMPASTV account. Comments taken using the first comment typed are not a reply to a comment, the comment data is retrieved using the Integrated Development Environment (IDE) Google Apps Script. After that, labeling for the comment data is carried out, and text preprocessing is carried out using several methods such as, Cleaning, case folding, tokenization, normalization, stopword removal, and stemming, then the words are given weights using TF-IDF (Term Frequency - Inverse Document Frequency). After that, feature selection was carried out using PSO (Particle Swarm Optimization) for up to 100 iterations and the number of features used was 933 features, then to continue making the classification model, sampling was carried out using SMOTE, which initially 645 negatif data, and 354 positive, became balanced 645 negatif, and 645 positive, the classification results obtained with an accuracy of 82.5%, a precision value of 78.7%, and a recall value of 89.1%.