Komparasi Metode BERT, VADER, dan RoBERTa untuk Analisis Sentimen Masyarakat terhadap Keputusan Pasangan

This research discusses the phenomenon of childfree in Indonesia, which is increasingly being discussed in line with social and economic changes. Although a negative stigma is still attached to a couple's decision not to have children, public awareness of the childfree option continues to increase. This study aims to analyze public sentiment towards childfree decisions using three sentiment analysis methods, namely BERT, RoBERTa, and VADER. The analysis results show that the BERT method has the highest accuracy of 99%, signaling its ability to classify sentiment very accurately. In contrast, the RoBERTa and VADER methods show lower accuracy, at 50% and 41% respectively. Both methods had difficulty in distinguishing the sentiment classes, which resulted in many misclassifications. Evaluation using the confusion matrix shows that RoBERTa and VADER have a significant number of misclassifications, with RoBERTa having 9 FPs and 19 FNs, and VADER having 16 FPs and 84 FNs. Meanwhile, BERT has almost no errors in classification, with a total FP of 0 and FN of 1. These results confirm that the BERT method is superior for sentiment analysis of the childfree phenomenon compared to the RoBERTa and VADER methods. This research provides insight into how people view the childfree phenomenon and finds the best sentiment analysis method among the three methods.