Comparative Study of Naïve Bayes Classifier and Support Vector Machine Methods in Public Sentiment Analysis of Prabowo-Gibran's Free Lunch Program

Abstract

In today's digital era, social media has become the main platform for people to voice their opinions on social and political issues. One of the most discussed topics is the free lunch program of President-elect Prabowo Subianto and Vice President-elect Gibran Rakabuming Raka. The program triggered various public reactions, making it relevant for sentiment analysis. The purpose of this study is to compare the performance of two text classification algorithms-Naïve Bayes and Support Vector Machine (SVM)-in classifying public sentiment towards the program. The dataset was obtained from Kaggle, with 657 initial data. After preprocessing, 156 data remained, consisting of 127 negative sentiments and 31 positive sentiments. Data processing followed the CRISP-DM framework, with Python and Scikit-learn used in model training. The results showed that the naive bayes classifier performed better with 84.38% accuracy, 86.90% precision, and 84.38% recall. Support Vector Machine showed lower performance in all metrics. In addition, the Naive Bayes Classifier was able to classify sentiments in a more balanced manner. The analysis was performed using Jupyter Notebook, and the final model was implemented through a Streamlit-based web interface.In today's digital era, social media has become the main platform for people to voice their opinions on social and political issues. One of the most discussed topics is the free lunch program of President-elect Prabowo Subianto and Vice President-elect Gibran Rakabuming Raka. The program triggered various public reactions, making it relevant for sentiment analysis. The purpose of this study is to compare the performance of two text classification algorithms-Naïve Bayes and Support Vector Machine (SVM)-in classifying public sentiment towards the program. The dataset was obtained from Kaggle, with 657 initial data. After preprocessing, 156 data remained, consisting of 127 negative sentiments and 31 positive sentiments. Data processing followed the CRISP-DM framework, with Python and Scikit-learn used in model training. The results showed that the naive bayes classifier performed better with 84.38% accuracy, 86.90% precision, and 84.38% recall. Support Vector Machine showed lower performance in all metrics. In addition, the Naive Bayes Classifier was able to classify sentiments in a more balanced manner. The analysis was performed using Jupyter Notebook, and the final model was implemented through a Streamlit-based web interface.