Comparative Analysis of Naïve Bayes Classifier and Support Vector Machine for Multilingual Sentiment Analysis: Insights from Genshin Impact User Reviews

This study evaluates the performance of Support Vector Machine (SVM) and Naïve Bayes Classifier (NBC) algorithms in analyzing user sentiments regarding the popular online game Genshin Impact. The dataset comprises reviews sourced from the Google Play Store and App Store in both Indonesian and English, reflecting linguistic diversity. Preprocessing techniques such as tokenization, stemming, and Term Frequency-Inverse Document Frequency (TF-IDF) were employed to enhance data quality. Sentiments were classified as positive, neutral, or negative using TextBlob-assisted processes. Results demonstrate that NBC outperformed SVM across all metrics, with an average accuracy of 71% compared to 63%. Notably, sentiment analysis on English datasets consistently achieved higher accuracy than on Indonesian datasets, emphasizing the challenges posed by the linguistic complexity of Indonesian. This research underscores the critical role of language-specific adaptations in improving machine learning algorithms for multilingual sentiment analysis. The findings provide actionable insights for optimizing user engagement through enhanced game feedback mechanisms.