Sentiment Analysis Related to Korean Subculture in Indonesia Using BERT Method

Abstract

Abstract

Korean subculture has experienced rapid growth in Indonesia in recent years. Indonesian society expresses positive, negative, or neutral views through social media X. (Twitter). Sentiment analysis is conducted to identify the public's views as positive, negative, or neutral. The stages of analysis include data collection, data pre-processing, data labeling, model training, classification, and visualization. The data analyzed consisted of 1,154,542 tweets from the years 2020 to 2023. The deep learning model BERT (IndoBERT) was used to classify the views of the Indonesian public obtained from social media X. A comparison of models with training-validation-testing data distribution parameters was conducted to determine the best model. The best model used a 70:20:10 data distribution with an accuracy of 93.60%, precision of 93.64%, recall of 93.60%, and an f1-score of 93.61%, as well as a validation accuracy of 92.70%. The analysis results show that sentiment towards Korean Subculture consists of 42.04% neutral, 32.24% positive, and 25.72% negative. Korean dramas are the most popular part of Korean subculture among the Indonesian public compared to music and food.