Penerapan Sistem Ensemble Learning untuk Analisis Sentiment Pembaruan One UI 7 pada Perangkat Samsung Menggunakan Algoritma Naive Bayes, SVM, LSTM, BI-LSTM, Dan IndoBERT

The One UI 7 interface update on Samsung devices has generated various user responses, including positive, negative, and neutral opinions. The large number of opinions shared on social media makes sentiment analysis a relevant approach to comprehensively understand user feedback. This study aims to analyze and evaluate the performance of an ensemble learning method using majority voting under two sentiment classification schemes, namely three-class sentiment classification (positive, negative, and neutral) and two-class sentiment classification (positive and negative). The data were collected from user comments on TikTok related to the One UI 7 update on Samsung devices, followed by labeling and text preprocessing stages. The classification models employed include Naïve Bayes, Support Vector Machine (SVM), Long Short-Term Memory (LSTM), Bidirectional Long Short-Term Memory (Bi-LSTM), and IndoBERT. The results of the study indicate that two-class sentiment classification achieves better performance, with the highest ensemble accuracy reaching 88.6% using the combination of Naïve Bayes, SVM, and IndoBERT. In comparison, three-class sentiment classification shows a lower performance, with the highest accuracy of only 75.8%, which is also obtained from the combination of Naïve Bayes, SVM, and IndoBERT.