ANALYSIS SENTIMENT ON VACCINE ISSUES IN INDONESIA USING NAÏVE BAYES AND MAXIMUM ENTROPY METHOD

• Main Author: Bertha Santigiovanni, Paulina
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/59608
• Institution: Institut Teknologi Bandung

Technological advances make people accustomed to conveying opinions, stories, news, experiences, and other things through online media with different sentiments. This sentiment can contain subjective statements that describe a person's perception of an event. Sentiment analysis is a natural language processing to find out the public's mood regarding a particular product or topic. The research in this Final Project aims to obtain sentiment analysis results using the Naïve Bayes and Maximum Entropy methods and compare the accuracy of the two methods to find out which method is better for sentiment analysis between the two. Sentiment analysis was carried out in stages from Twitter crawling using Python using the keyword "vaccine" in Indonesian, preprocessing data with RapidMiner. Then classification based on positive, negative, or neutral sentiment using the Naive Bayes method RapidMiner and Maximum Entropy in Python. Sentiment data was analyzed on vaccine issues in Indonesia through Twitter social media in two weeks. Based on the data processing results, an accuracy of 63.84% was obtained with the Naïve Bayes and an accuracy of 80.69% with the Maximum Entropy. Then, the amount of training data and the number of iterations were varied. Then it is obtained that more and more training data does not always increase accuracy. The conclusion obtained by the Maximum Entropy method produces higher accuracy and precision than Naïve Bayes in this study.