Urban Intel: AI-Based Public Issue Detection from Social Media for Municipal Decision Support

Rapidly growing cities such as Bengaluru face recurring civic problems, including potholes, waste accumulation, water leakage, blocked drains, traffic congestion and broken streetlights. Citizens increasingly describe these problems on social media, but useful reports are mixed with unrelated conversation, reposts, jokes, incomplete descriptions and duplicates, and most of them never reach municipal authorities. UrbanIntel is a web-based public issue detection and civic intelligence system that organizes this information into a reviewable workflow. Its pipeline collects candidate public posts, validates their source links, pre-processes the text, checks civic relevance, classifies the issue with a three-stage hybrid approach (keyword matching, pattern recognition and semantic scoring), extracts location clues, screens for duplicates and presents structured records through a dashboard and a citizen-facing interface. The frontend is built with React and Vite, with Recharts charts and Leaflet maps; the backend is a Node.js and Express service, and Reddit is the current social-media source. The system is intended to support, not replace, municipal verification and official complaint channels. This paper documents the problem, objectives, related work, architecture, methodology, implementation, testing approach, limitations and future enhancements. Quantitative accuracy and performance results are not claimed, because a verified labelled evaluation dataset and measured results are not yet available.