<![CDATA[Digital transformation makes the world change rapidly, especially in the development of metaverse technology. The development of metaverse technology has received positive and negative responses from the public, so it is necessary to analyze whether public opinion accepts the development of metaverse technology or vice versa. This research aims to analyze 6728 public comment data regarding the metaverse on social media X using a text mining approach. By comparing text mining algorithm models, this experiment seeks to find the best algorithm for metaverse sentiment analysis, thereby providing insight to industry players involved in metaverse development. This research experiment uses a comparison of two algorithms, namely Naïve Bayes and Logistic Regression. The comparison results for the Naïve Bayes algorithm have an accuracy value of 90% and Logistic Regression of 91%, but the precision, recall, and F1-Score results are low. This indicates that the machine predominantly learns positive sentiment because this sentiment has a majority label, namely 5799 positive sentiment data, while negative sentiment is a minority label with 795 data. To overcome the problem of unbalanced data (Imbalance) in this research, SMOTE optimization was used. The results of SMOTE optimization have a superior value in the Logistic Regression algorithm, the accuracy value of 95% has also increased in the confusion matrix, namely the precision value of 94%, recall of 93%, and F1-Score of 95%. Meanwhile, the Naïve Bayes algorithm has a smaller value, namely 91% accuracy, and the negative sentiment confusion matrix has increased to 87% precision, 97% recall, and 92% F1-Score, so the accuracy and confusion matrix values have better performance.]]>
