Urban traffic congestion continues to pose serious challenges to mobility, energy efficiency, and environmental sustainability. To address this problem, this paper proposes a deep Q-Network (DQN)-based real-time recommendation model to manage dynamic traffic signals across multiple intersections. The proposed model integrates reinforcement learning with a data-driven recommendation engine within the Simulation of Urban Mobility (SUMO) environment, utilizing traffic attributes such as queue length, vehicle speed, waiting time, and throughput. A novel reward function, which combines queue minimization and wait time reduction, significantly enhances traffic flow efficiency. The experimental results reveal substantial improvements in average speed, queue length, and aggregate waiting time relative to baseline models. This paper demonstrates the feasibility and effectiveness of DQN for real-time, multi-intersection traffic signal management and its use in recommendation system.