Machine Learning in urban traffic monitoring and demand prediction: systematic review.

Resumen

Rapid urban growth and vehicular congestion have driven the search for technological solutions to improve mobility in urban environments. This systematic review analyzed Machine Learning techniques applied to traffic monitoring and passenger demand prediction. Different approaches used in recent literature were explored, as well as their effectiveness compared to traditional methods. The study was conducted following the PICO strategy and PRISMA methodology to ensure a rigorous and transparent process. A systematic search was carried out in scientific databases such as Scopus, IEEE Xplore and Web of Science, initially identifying 660 articles. Using the PICO strategy, the general question and keywords that helped the search for articles were generated. After applying the inclusion and exclusion criteria defined in the PRISMA protocol, 47 relevant studies were selected between 2021 and 2025.The findings were organized to answer four research questions: definition of the urban transportation system, techniques used, effectiveness versus traditional methods, and improvements achieved. It was identified that the most commonly employed techniques included deep neural networks, reinforcement learning, and hybrid models. These tools were applied to tasks such as vehicle flow prediction, incident detection and transportation user behavior analysis. In general, Machine Learning-based models demonstrated significant advantages in accuracy, adaptive capacity and real-time data processing, especially in dynamic urban contexts.It was concluded that Machine Learning represents a key tool for the transformation of urban transport systems, contributing to more efficient, resilient and data-driven planning.