Traffic prediction in time series, spatialtemporal, and OD data: A systematic survey
Kai Du, Xingping Guo, Letian Li, Jingni Song, Qingqing Shi, Ming Hu, Jianwu Fang
Abstract
The burgeoning field of intelligent transportation systems (ITS) has been pivotal in addressing contemporary traffic challenges, significantly benefiting from the evolution of computational capabilities and sensor technologies. This surge in technical advancement has paved the way for extensive reliance on deep-learning methodologies to exploit large-scale traffc data. Such efforts are directed toward decoding the intricate spatiotemporal dynamics inherent in traffc prediction. This study delves into the realm of traffc prediction, encompassing time series, spatiotemporal, and origin-destination (OD) predictions, to dissect the nuances among various predictive methodologies. Through a meticulous examination, this paper highlights the efficacy of spatiotemporal coupling techniques in enhancing prediction accuracy. Furthermore, it scrutinizes the existing challenges and delineates open and new questions within the traffc prediction domain, thereby charting out prospective avenues for future research endeavors.