<![CDATA[This study aims to analyze public sentiment toward the policy of the Pusat Pelaporan dan Analisis Transaksi Keuangan (PPATK) regarding bank account blocking, as expressed on the social media platform X (formerly Twitter), using a machine learning approach. The research data were collected through a web crawling process that gathered 799 tweets using Tweet Harvest during the period of July 28 to August 1, 2025. This time frame was selected because discussions related to account blocking by PPATK experienced a significant surge and became one of the viral topics on social media X. The analytical stages consisted of text preprocessing, data visualization, sentiment labeling using VADER, feature extraction with Term Frequency–Inverse Document Frequency (TF-IDF), data splitting, class balancing using the Synthetic Minority Oversampling Technique (SMOTE), and the development of classification models employing Support Vector Machine (SVM) and Decision Tree (DT) algorithms. The results indicate that the majority of public opinion was negative, accounting for 92.1% of the data, while positive opinions comprised only 7.9%. Both SVM and DT achieved high accuracy levels ranging from 93% to 94%. However, after hyperparameter tuning using GridSearchCV, the Decision Tree model demonstrated more balanced performance in detecting the minority (positive) class, with a precision of 0.73, recall of 0.65, and an F1-score of 0.69. These findings suggest that the integration of comprehensive preprocessing, TF-IDF feature representation, SMOTE-based class balancing, and parameter tuning can significantly enhance sentiment classification performance. Furthermore, this study provides a comprehensive overview of public perceptions of government policy. The results highlight the importance of utilizing machine learning–based sentiment analysis as a strategic consideration in formulating public policies that are more responsive to societal aspirations.]]>
