<![CDATA[The leadership of Prabowo and Gibran has drawn significant public attention, particularly in the early phase of the administration. In the digital era, public opinions both supportive and critical are widely expressed through platforms such as YouTube, generating large volumes of data. Manual analysis of this data is impractical, while previous machine learning and deep learning approaches have produced suboptimal results. Therefore, this study applies a Deep Neural Network (DNN) to analyze public sentiment more effectively. The process involves inputting a dataset derived from YouTube comments obtained through scraping using the v3 API. This is followed by preprocessing, labeling, data splitting, applying IndoBERT word embedding, and DNN, with the final step being evaluation using a confusion matrix. The results of this study indicate that as the depth of the hidden layers increases, the accuracy improves, though not significantly. The best model performance was obtained by without SMOTE using a depth of 6 hidden layers, with an accuracy of 68.84%, precision of 67.5%, recall of 68.84%, and an F1 score of 64.22%.]]>
Copyrights © 2026
