Recent Advances in Traffic Forecasting Using Spatio-Temporal Networks and Machine Learning for Intelligent Transportation Systems

Smart transportation systems have transformed how we travel, making mobility more efficient by utilizing mobile internet and positioning technologies to gather spatio-temporal data. At the heart of Intelligent Transportation Systems (ITS) lies traffic prediction, which plays a crucial role in optimising transportation networks. This paper explores traffic prediction in depth, covering everything from the foundational ITS framework to the processing of spatio-temporal data. It breaks down the research landscape into five key areas: Spatio-Temporal Analytics, Data Preparation, Traffic Forecasting Challenges, Feature Selection and Extraction, and Urban Mobility Applications. Through a review of 235 studies published between 1986 and the present, this paper provides insights into current challenges and emerging trends in traffic prediction. It examines techniques for selecting and extracting relevant features from spatio-temporal data to enhance forecasting accuracy while maintaining reliability. Additionally, it highlights major obstacles in traffic forecasting, such as traffic segmentation, flow simulation, and demand modelling, while shedding light on innovative approaches to address them. The study also discusses real-world applications, including dispatch management, shared mobility services, anomaly detection, and itinerary mapping, offering solutions to improve urban transportation efficiency. By addressing gaps in previous research, this paper presents a fresh, up-to-date perspective on the field, helping researchers and professionals navigate the evolving landscape of intelligent transportation.