| Abstract |
Public aspiration complaints can be vast in number and varied in content. In Indonesia, there are online-based public complaint service platforms provided by the government regarding public policies and services. This public aspiration data involves large and complex datasets specially in Indonesian. Advanced machine learning algorithms can achieve high accuracy in sentiment classification by learning from this public aspiration complaint data. Aspect-Based Sentiment Analysis (ABSA) is a Natural Language Processing (NLP) technique used to analyze and evaluate sentiments expressed towards specific aspects or attributes in a text. ABSA can efficiently perform this multi-layered analysis, which is more complex than general sentiment analysis. In this research, ABSA will be conducted using machine learning methods. The dataset used is obtained from the public aspiration complaint website, ulas.surakarta.go.id. The machine learning methods utilized in this research are Support Vector Machine (SVM), Complement Naive Bayes, and Decision Tree. The results show that the best model in this study is the SVM algorithm, which yielded an accuracy of 85.65%, outperforming the Complement Naive Bayes, which yielded an accuracy of 81.0%, and the Decision Tree, which yielded an accuracy of 82.77%. The public aspiration complaints (pengaduan) aspect has the most negative sentiment, particularly regarding issues in educational assistance, school education, family programs, and policies related to job industries. For other aspects, such as requests (permohonan), questions (pertanyaan), and suggestions (usulan), people tend to respond neutrally. These findings can expedite decision-making processes, helping to prioritize issues that matter most to the public towards smart city. © 2024 IEEE. |
