Following the emergence of the knowledge-based economy, the digital economy, and the intelligent economy, smart cities are poised to represent the next phase in urban development. These cities aim not only to leverage both physical and digital infrastructures for urban advancement but also to harness intellectual and social capital as essential elements of urbanization. Smart cities are recognized as pivotal catalysts for transforming various sectors by integrating multiple municipal systems such as transportation, healthcare, and operational frameworks. The concept of a smart society evolves from smart cities, characterized by a digitally interconnected, knowledge-driven community that actively pursues social, environmental, and economic sustainability. Recently, deep learning has gained traction due to its ability to effectively tackle complex problems across diverse applications using both supervised and unsupervised learning methods. This approach relies on advanced techniques for managing large datasets and multilayer neural networks, which often outperform traditional ANN in processing historical data. This paper introduces a novel algorithm based on deep learning designed to accurately predict traffic flow behavior. The algorithm learns from multivariate sequence data by analyzing spatio-temporal dependencies and non-linear correlations. Simulation results demonstrate that the proposed method surpasses existing algorithms in performance.