Machine Learning Models for Traffic Congestion Prediction in Urban Smart Mobility Systems

Urban traffic congestion poses significant challenges to sustainable mobility, economic productivity, and environmental quality in modern cities. With the increasing deployment of Internet of Things (IoT) devices and real-time data collection systems, smart cities are generating vast volumes of traffic-related data that can be harnessed for predictive analytics. This study carefully looks at how machine learning models are used to accurately predict traffic jams in smart city transportation systems. We check and compare how well different supervised learning models work, like random forest networks, support vector regression, gradient boosting machines, and long short-term memory, using real traffic data from sensors, GPS devices, and city infrastructure. The process involves cleaning up data, creating features, training models, and adjusting settings to get the best prediction results. The tests showed that LSTM networks, because they can understand patterns over time, are better than traditional machine learning at predicting traffic jams, with a root mean squared error of 5.40 and a mean absolute percentage error of 9.7%. However, tree-based models like GBM are good for providing clear and efficient explanations, along with good accuracy, making them useful in smart city environments with limited resources. This research paper discusses the importance of understanding how the model works, choosing the right features, and the effects of adding machine learning-based prediction tools to city traffic management systems. The findings support the idea that machine learning can greatly improve real-time traffic monitoring, allowing for quick action to reduce its effects and improve smart city transportation.