Komparasi Naïve Bayes dan SVM Analisis Sentimen RUU Kesehatan di Twitter

Abstract

This research focuses on sentiment analysis regarding the plan to ratify the Health Bill which has become a hot topic of conversation on social media, especially Twitter. This research aims to classify tweets that reflect various opinions regarding the Health Bill, including support, rejection and neutrality. In this research, the author uses two types of classification algorithms, namely the Multinomial Naïve Bayes Algorithm and the Support Vector Machine (SVM) Algorithm. Previously, tweets were labelled using the Lexicon InSet dictionary. The research was conducted in the Python programming language and using Google Collaboratory. Data validation was carried out using the K-fold cross-validation method. The research results indicate that both algorithms predominantly produce positive sentiments over negative ones. However, SVM with a linear kernel achieves a higher accuracy rate of 0.87, compared to Multinomial Naïve Bayes, which has an accuracy of 0.82. SVM also records a precision of 0.87, recall of 0.97, and an F1-score of 0.91, while Multinomial Naïve Bayes shows a precision of 0.81, recall of 0.98, and an F1-score of 0.89. Overall, this research confirms that SVM excels in text sentiment classification, while Multinomial Naïve Bayes also provides good results in recognising positive and negative sentiment. These results have important implications for understanding public sentiment regarding the Health Bill on the Twitter platform.