Sentiment analysis of public sensitivity to COVID-19 vaccines on Twitter by majority voting classifier-based machine learning
Yükleniyor...
Dosyalar
Tarih
2023
Yazarlar
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
Gazi University, Faculty of Engineering and Architecture
Erişim Hakkı
info:eu-repo/semantics/openAccess
Özet
Purpose: The aim of this study is to analyze public sentiment with Machine Learning of vaccine-related tweets obtained on Twitter in order to better understand the attitudes and concerns of social media users, especially regarding COVID-19 vaccines in Turkey. For this purpose, a majority voting method has been developed with machine learning methods, which are frequently used in sentiment analysis studies.Theory and Methods: In the study, machine learning algorithms used in six different classification tasks, which are frequently used in sentiment analysis, were compared. Then, by comparing these machine learning methods, the majority voting method, which is an ensemble learning method, was developed by using the three methods with the highest accuracy. For this purpose, both soft voting and hard voting methods were used to generate majority voting in the classification task. In addition, the data used in the study were collected between 01.04.2021 and 31.08.2021, when vaccine studies accelerated in Turkey, a total of 412,588 tweets in Turkish. Results: Although the SVM algorithm among the individual methods achieved a high success rate of 89.6%, it is seen that the XGBoost model is the most successful algorithm with a accuracy rate of 89.8%. Although the Random Forest approach among other machine learning approaches has achieved remarkable success, the same is not the case for other methods. For this reason, high accuracy SVM, XGBoost and Random Forest methods are used in both hard voting and soft voting majority voting approaches. Although the hard voting method achieved a higher accuracy than the individual methods with a success rate of 88.9%, the soft voting method was the most successful classification method with a relatively high accuracy rate of 90.5%. For this reason, soft voting approach was used in the labeling of daily tweets obtained in the study.Conclusion: As a result of the analyses carried out with the soft voting method, although there are fluctuations in the sentiment polarity of the tweets about the vaccine, it is observed that the negative sentiments and therefore the opposition to the vaccine is becoming more and more popular on social media. Particularly, when compared to previous study findings, positive sentiments in vaccine-related posts in Turkey are quite low rate. For this reason, the ongoing opposition to vaccination on social media in Turkey becomes a subject that needs to be examined more carefully. As far as we know, this study is the first in Turkey to perform sentiment analysis on COVID-19 vaccines. In addition, the findings of the study show that the proposed method is a valuable and easily applied tool to monitor the sensitivity of COVID-19 vaccines with a sentiment analysis approach via social media.
Açıklama
Anahtar Kelimeler
COVID-19, Vaccine, Sentiment Analysis, Majority Voting Classifier, Machine Learning, Text Classification
Kaynak
Journal of the Faculty of Engineering And Architecture of Gazi University
WoS Q Değeri
Q3
Scopus Q Değeri
Q2
Cilt
38
Sayı
2
Künye
Çılgın, C., Gökçen, H., & Gökşen, Y. (2023). Sentiment analysis of public sensitivity to COVID-19 vaccines on Twitter by majority voting classifier-based machine learning. Journal of the Faculty of Engineering and Architecture of Gazi University, 38(2).
