Comparison of Support Vector Machine and Naïve Bayes to Sentiment Analysis of Military Barracks Program

Abstract

Sentiment analysis is a study that analyzes a person's emotions about a problem. The military barracks program proposed by the Governor of West Java has drawn pros and cons from the community, especially in application X. Some people consider this program to be the right solution to discipline and shape the character of students, others think that the program can take away children's freedom and rights and does not guarantee any change in character after the students leave the military barracks. Therefore, a sentiment analysis was conducted with the aim of understanding public sentiment and comparing the accuracy of the SVM and Naïve Bayes in predicting public sentiment towards the military barracks program. The method in this study begins with data crawling, data selection, labeling, data preprocessing (data cleaning, normalization, case folding, stopword removal, tokenizing, stem), TF-IDF, Word Cloud, classification with Naïve Bayes and SVM, ending with a Confusion Matrix. In contrast to SVM, which revealed that 1429 tweets had positive sentiment and 447 had negative sentiment, Naïve Bayes results indicated that 1309 tweets had positive sentiment and 567 had negative sentiment. The accuracy value of the Naïve Bayes was 91.24%, the precision was 99.73%, and the recall was 82.94%. In contrast, the SVM achieved 92.16%, the precision was 97.86%, and the recall was 86.40%. Based on these findings, it can be said that the SVM  is more accurate than Naïve Bayes and that the public generally has a favorable opinion of the military barracks program.