Analisis Sentimen Pengguna Coretax Pada Media Sosial Menggunakan Decision Tree Dan Adaptive Boosting

The implementation of Indonesia's latest digital tax administration system, Coretax, has sparked various opinions and complaints from the public on social media. The large volume of unstructured public comments makes it difficult for authorities to evaluate services manually. This study aims to perform sentiment analysis on Coretax users on X (Twitter) and Instagram platforms to classify opinions into positive, negative, and neutral categories. The method employed is the Decision Tree algorithm optimized with the Adaptive Boosting (AdaBoost) ensemble learning technique. The research stages include web scraping, text preprocessing (including negation merging), word weighting using TF-IDF, and model evaluation. The dataset consists of 1,414 records after undergoing an augmentation process to maintain class balance. The results show that AdaBoost optimization successfully increased the model's accuracy by 13.78% from the Decision Tree baseline, reaching a final accuracy of 86.22%. Furthermore, the system is implemented into a web-based application using a hybrid Laravel-Flask architecture, allowing for interactive visualization of analysis results. This research is expected to serve as a tool for the Directorate General of Taxes (DJP) to understand user complaint patterns and improve the quality of digital tax services in real-time. Keywords: Sentiment Analysis, Coretax, Decision Tree, Adaptive Boosti