AI-Driven Flood Prediction and Mitigation System for Sustainable Cities: Integrating Google ML Kit and Gemini API for Smart Assistance

Floods are among Malaysia’s most frequent and disruptive natural disasters, particularly in Johor during the Northeast Monsoon season. Current flood management strategies suffer from centralisation, delayed alerts, and ineffective resource coordination, leaving communities vulnerable. This project introduces an AI‑Driven Flood Prediction and Mitigation System that integrates machine learning, real-time data sharing, and intelligent resource management within a mobile and web-based platform. Developed using the Agile Unified Process (AUP), the system comprises a Flutter-based mobile app and a Django-administered web portal. Key features include on-device IC verification via Google ML Kit, AI-powered emergency checklists using the Gemini API, and GPS-tagged incident reporting. The platform enables seamless victim registration, crowdsourced data collection, and dynamic coordination between users, rescuers, NGOs, and administrators. Real-time analytics and predictive models, informed by live river data and historical trends, enhance situational awareness and early-warning capabilities. The system’s design was validated through technical testing and user desirability evaluations, achieving high ratings for usability (4.58/5) and innovation (4.67/5). Automated data ingestion, context-aware AI support, and QR-based group identification contribute to a streamlined disaster response ecosystem. By aligning with Sustainable Development Goal 11: Sustainable Cities and Communities, this initiative offers a scalable framework for proactive and community-driven flood resilience. Future developments aim to extend support to multilingual users and incorporate IoT-based flood forecasting technologies.