A Systematic Review of Machine Learning Approaches for K-Pop Concert Sentiment Analysis on X

K-pop concerts in Indonesia generate intensive digital discussion on X/Twitter, yet studies that directly combine concert-related objects, X data, and machine-learning-based sentiment classification remain limited. This study conducts a systematic literature review to map research objects, methods, preprocessing techniques, evaluation results, and research gaps in K-pop sentiment analysis. The selection process followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses framework. A total of 5,674 records were collected, 4,618 duplicates were removed, 1,056 records were screened, and 20 articles were included in the final synthesis. The findings show that X is the dominant platform. Previous studies more frequently examine K-pop groups, fandom, the Korean Wave, and cyberbullying rather than direct concert experiences. Naive Bayes remains widely used because it is simple, efficient, and suitable for high-dimensional text data, although Support Vector Machine and transformer-based models often provide stronger performance in specific settings. Classification quality is strongly affected by non-standard language normalization, multilingual content, class balance, feature weighting, and labeling consistency. The main gap is the absence of an Indonesian K-pop concert sentiment-analysis design that combines domain-aware preprocessing, per-class evaluation, and aspect-level interpretation of ticketing, promoters, venues, safety, and audience experience.