This paper presents a novel adaptive traffic signal control framework that integrates Graph Neural Networks (GNNs), Multi-Agent Reinforcement Learning (MARL), and real-time crowdsourced event data from Waze to address dynamic urban traffic challenges. The proposed system leverages spatial-temporal graph modeling to predict short-term traffic states, enabling decentralized agents at individual intersections to optimize signal timing policies collaboratively. An event-triggered adaptation mechanism allows rapid policy adjustments in response to incidents such as accidents or roadblocks, enhancing traffic flow resilience. Extensive simulations using real-world traffic data from Ho Chi Minh City demonstrate that the framework significantly reduces vehicle waiting times by over 25%, increases average speeds, and substantially decreases traffic deadlocks compared to fixed-time control strategies. These results underscore the potential of combining graph-based deep learning with reinforcement learning and crowdsourced event integration for effective, scalable urban traffic management.