Perbandingan Metode Naïve Bayes dan Support Vector Machine Untuk Analisis Sentimen Terhadap Vaksin Astrazeneca di Twitter

Abstract

The implementation of Covid-19 vaccination in Indonesia turned out to have various pro and contra opinions from the public. The discovery of disinformation and misinformation about vaccines spread through social media content affects a person's absorption of information so which leads to vaccine delays. When in fact, vaccination is one of the biggest and most effective contributions to preventing the Covid-19 pandemic. Astrazeneca is one of the vaccines provided by the Indonesian government. This vaccine used to be controversial amongst the public regarding its halalness and the safety of the vaccine because of the issue of the said vaccine containing swine trypsin. Nowadays Twitter has become a place for users to express their concerns and opinion regarding the Covid-19 vaccine. Data obtained from Twitter will be useful if it is analyzed, one of which is sentiment analysis. In this study, data collection was carried out using the snscrape library with a total of 3105 tweets obtained from the period May to June 31, 2021. The dataset that has been collected is then preprocessed to optimize the data. After passing the preprocessing stage, the data was labeled as tweet class using a lexicon-based dictionary which resulted in 1275 tweets with positive opinin labels and 1830 tweets labeled as negative opinion. The aim of this study is to examines the performance of Naïve Bayes and Support Vector Machine with adding the weighting method TF-IDF (Term Frequency – Inverse Document Frequency). The evaluation results show that the Support Vector Machine has a greater accuracy, precision, recall and f1-score of 87.27%, 90.41%, 77,34% and 83.37% compared to Naïve Bayes which has an accuracy, precision, recall and f1- of 76.81%, 72.40%, 70.70% and 71.52%.