Urbanization has significantly exacerbated traffic congestion and air pollution, posing severe environmental and public health challenges. This study introduces an integrated framework that leverages machine learning models, including Seasonal Autoregressive Integrated Moving Average (SARIMA), Support Vector Machines (SVM), and Long Short-Term Memory (LSTM), to predict pollutant levels and monitor traffic congestion. By employing real-time sensor data and predictive analytics, the system identifies critical congestion levels based on emissions of carbon monoxide (CO) and nitrogen oxides (NOx). The analysis demonstrates the superior performance of SARIMA in forecasting long- term trends, while LSTM effectively captures dynamic fluctuations. This interdisciplinary approach provides scalable solutions for urban pollution management, emphasizing the potential of artificial intelligence to enable sustainable smart city ecosystems.