Perbandingan Naïve Bayes Classifier dan Support Vector Machine Terhadap Sentimen Vaksinasi Covid-19 di Indonesia

Amalia, Dini and Dwi Ayu Lusia, S.Si., M.Si. (2022) Perbandingan Naïve Bayes Classifier dan Support Vector Machine Terhadap Sentimen Vaksinasi Covid-19 di Indonesia. Sarjana thesis, Universitas Brawijaya.

Abstract

Naïve Bayes Classifier (NBC) dan Support Vector Machine (SVM) merupakan metode klasifikasi yang sering digunakan dalam mengklasifikasikan sentimen masyarakat terhadap suatu fenomena tertentu. Pada 13 Januari 2021, pemerintah mencanangkan program vaksinasi untuk meminimalisir penyebaran virus covid-19. Tujuan penelitian ini yaitu untuk membandingkan hasil klasifikasi algoritma NBC dan SVM, dengan topik sentimen publik terkait vaksinasi covid�19 di Indonesia dalam media sosial twitter. Pengumpulan data dari tanggal 13 Januari – 31 Desember tahun 2021 diperoleh sebanyak 288.482 tweet. Sentimen yang didapatkan berupa 10.447 sentimen positif, 5.677 sentimen negatif, dan 272.358 sentimen netral. Setelah sentimen netral dihilangkan, hasil pelabelan data menunjukkan bahwa pandangan masyarakat terkait topik tersebut cenderung positif dengan opini sebesar 65% tweet dan 35% sisanya berisi komentar negatif. Selanjutnya dilakukan tahapan preprocessing untuk memperbaiki kata yang belum tepat dalam tweet. Kemudian, data dibagi menjadi data training dan data testing dengan perbandingan 80%:20%. Ketepatan klasifikasi yang dipilih untuk membandingkan kedua metode adalah accuracy, recall, precision, dan f1-score. Klasifikasi SVM dalam penelitian ini memiliki performa yang lebih baik dibandingkan NBC. Hasil accuracy, recall, precision, dan f1-score pada SVM berturut�turut adalah 96,5892%, 96,5892%, 96,6225%, dan 96,5659%, sedangkan NBC menghasilkan performa sebesar 84,9302%, 84,9302%, 87,9218%, dan 85,2640%. Berdasarkan hal tersebut, disimpulkan bahwa SVM dapat digunakan sebagai algoritma klasifikasi terkait sentimen publik pada twitter dalam penelitian selanjutnya.

English Abstract

Naïve Bayes Classifier (NBC) and Support Vector Machine (SVM) are classification methods often used in classifying public sentiment towards a particular phenomenon. On January 13th 2021, the government launched a vaccination program to minimize the spread of the covid-19 virus. The purpose of this study is to compare the results of the NBC and SVM algorithm classifications, with the topic of public sentiment related to covid-19 vaccination in Indonesia on twitter. Data collection from January 13th to December 31st 2021 obtained 288.482 tweets. The sentiment obtained was 10.447 positive sentiments, 5.677 negative sentiments, and 272.358 neutral sentiments. After neutral sentiment was eliminated, the data labeling results showed that people's views on the topic tended to be positive with 65% of tweets containing positive comments and the remaining 35% containing negative comments. Furthermore, the preprocessing stage is carried out to correct the word that is not appropriate yet in the tweet. Then, the data is divided into training data and testing data with 80%:20% ratio. The accuracy of the classification chosen to compare the two methods is accuracy, recall, precision, and f1-score. The SVM classification in this study performed better than NBC. Accuracy, recall, precision, and f1-score results on SVM were 96.5892%, 96.5892%, 96.6225%, and 96.5659%, respectively, while NBC produced performance of 84.9302%, 84.9302%, 87.9218%, and 85.2640%. Based on this, it was concluded that SVM can be used as a classification algorithm related to public sentiment on Twitter in future research.

Item Type:	Thesis (Sarjana)
Identification Number:	052309
Uncontrolled Keywords:	Covid-19 Vaccination, Naïve Bayes Classifier, Sentiment Analysis, Support Vector Machine, Twitter
Divisions:	Fakultas Matematika dan Ilmu Pengetahuan Alam > Statistika
Depositing User:	Unnamed user with email prayoga
Date Deposited:	19 Jan 2024 08:54
Last Modified:	19 Jan 2024 08:54
URI:	http://repository.ub.ac.id/id/eprint/212776

Text (DALAM MASA EMBARGO) Dini Amalia.pdf Restricted to Registered users only until 31 December 2025. Download (3MB)

Actions (login required)

View Item