SENTIMENT ANALYSIS RELATED TO DIGITAL BANK SERVICES IN INDONESIA USING NAïVE BAYES AND MAXIMUM ENTROPY METHODS

• Main Author: Ivana Yuwono, Clarissa
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/75212
• Institution: Institut Teknologi Bandung

Physics is a part of science that essentially is a collection of knowledge, ways of thinking, methods for investigating the nature of the universe, as well as interactions with technology and society. Interactions between technology and society can also be referred to as complex systems, which have been applied to machine learning and artificial intelligence. Technological advancements that integrate all aspects have brought about significant changes, including in the banking industry. The consumption patterns of society shifting towards digital have prompted banks to undergo transformation by developing digital banking services in Indonesia. Technological progress has also made it easier for people to obtain information and express opinions through social media platforms like Twitter. This research aims to analyze the sentiment regarding digital banking services in Indonesia using the Naïve Bayes and Maximum Entropy methods. The analysis is conducted in several steps, starting with data crawling on Twitter using RapidMiner with keywords such as "bank digital," "bank jago," "neobank," "bank jenius," "seabank," and "blu bca." This is followed by data preprocessing and sentiment classification into positive or negative using the Naïve Bayes method in RapidMiner and the Maximum Entropy method using Python. Based on the data processing results, an accuracy of 82.98% was obtained in the first stage, and 58.89% in the second stage using the Naïve Bayes method, while with the Maximum Entropy, an accuracy of 85.11% was achieved in the first stage and 90.44% in the second stage. Furthermore, variations in the number of training data were explored, and it was found that increasing the number of training data does not always lead to an improvement in accuracy. The conclusion drawn from this research is that the Maximum Entropy method yields higher accuracy and precision compared to the Naïve Bayes method in this study.