Integration of Sentiment Analysis and Anime Character Popularity Metrics Using Natural Language Processing

Abstract

This study proposes a Natural Language Processing (NLP)-based sentiment analysis approach to examine the relationship between audience opinions and anime character popularity using an Indonesian-language review corpus, which remains underexplored in previous research. Employing TF-IDF representation and Logistic Regression, the study not only classifies sentiment polarity but also evaluates how sentiment contributes to variations in character popularity. A total of 557 reviews were analyzed, and the developed model achieved an accuracy of 92.15%, outperforming comparative models such as SVM and Naïve Bayes. The main contributions of this research are: (1) the utilization of an Indonesian-language dataset for anime sentiment analysis, addressing a notable gap in existing literature; (2) empirical evidence of the sentiment–popularity relationship grounded in audience perception theory and digital engagement behavior; and (3) a comparative model evaluation that strengthens methodological reliability. The results offer practical implications for the digital entertainment industry, including sentiment-driven recommendation systems, character popularity telemetry, and audience preference analytics to support strategic content development and promotional decisions.