Analysis of Public Sentiment Towards the Use of AI in Monitoring Waste via the SEMAR Monitoring Web and Its Impact on Flood Management in Semarang City

Abstract

Waste management and flood mitigation are key challenges in Semarang City. The Semarang City Government, through the Department of Communication and Informatics (Diskominfo), implemented the AI-based Pantau Semar web system to automatically detect waste accumulation and water puddles via a CCTV network. This study examines public sentiment toward the system, particularly from social media comments, and develops a sentiment classification model using IndoBERT. A quantitative approach with Natural Language Processing (NLP)-based sentiment analysis was applied. A total of 430 public comments from social media were classified into positive, negative, and neutral sentiments, and analyzed using a fine-tuned IndoBERT model. Results show that negative sentiment dominates (54%), followed by positive (30%) and neutral (16%). The model achieved 81% accuracy, with the highest F1-score in the negative class (0.89). These findings indicate that the public remains critical of the system’s performance, especially regarding waste accumulation and flooding, while also highlighting AI’s potential in environmental management and public opinion detection. The results provide a basis for developing more adaptive monitoring systems and improving government communication strategies to better address community needs.