ADST: Forecasting Metro Flow Using Attention-Based Deep Spatial-Temporal Networks with Multi-Task Learning

doi:10.3390/s20164574

	ADST: Forecasting Metro Flow Using Attention-Based Deep Spatial-Temporal Networks with Multi-Task Learning
	Jia, Hongwei 1; Luo, Haiyong 2; Wang, Hao 1; Zhao, Fang 1; Ke, Qixue 1; Wu, Mingyao 1; Zhao, Yunyun 1
	2020-08-01
发表期刊	SENSORS
卷号	20 期号:16 页码:23
摘要	Passenger flow prediction has drawn increasing attention in the deep learning research field due to its great importance in traffic management and public safety. The major challenge of this essential task lies in multiple spatiotemporal correlations that exhibit complex non-linear correlations. Although both the spatial and temporal perspectives have been considered in modeling, most existing works have ignored complex temporal correlations or underlying spatial similarity. In this paper, we identify the unique spatiotemporal correlation of urban metro flow, and propose an attention-based deep spatiotemporal network with multi-task learning (ADST-Net) at a citywide level to predict the future flow from historical observations. ADST-Net uses three independent channels with the same structure to model the recent, daily-periodic and weekly-periodic complicated spatiotemporal correlations, respectively. Specifically, each channel uses the framework of residual networks, the rectified block and the multi-scale convolutions to mine spatiotemporal correlations. The residual networks can effectively overcome the gradient vanishing problem. The rectified block adopts an attentional mechanism to automatically reweigh measurements at different time intervals, and the multi-scale convolutions are used to extract explicit spatial relationships. ADST-Net also introduces an external embedding mechanism to extract the influence of external factors on flow prediction, such as weather conditions. Furthermore, we enforce multi-task learning to utilize transition passenger flow volume prediction as an auxiliary task during the training process for generalization. Through this model, we can not only capture the steady trend, but also the sudden changes of passenger flow. Extensive experimental results on two real-world traffic flow datasets demonstrate the obvious improvement and superior performance of our proposed algorithm compared with state-of-the-art baselines.
关键词	forecasting passenger flow spatiotemporal networks multi-task learning attention mechanism spatiotemporal dependency
DOI	10.3390/s20164574
收录类别	SCI
语种	英语
资助项目	National Key Research and Development Program[2018YFB0505200] ; Action Plan Project of the Beijing University of Posts and Telecommunications ; Fundamental Research Funds for the Central Universities[2019PTB-011] ; Fundamental Research Funds for the Central Universities[2019XD-A06] ; Special Project for Youth Research and Innovation, Beijing University of Posts and Telecommunications ; National Natural Science Foundation of China[61872046] ; Joint Research Fund for Beijing Natural Science Foundation and Haidian Original Innovation[L192004] ; Key Research and Development Project from Hebei Province[19210404D] ; Key Research and Development Project from Hebei Province[20313701D] ; Science and Technology Plan Project of Inner Mongolia Autonomous Regio[2019GG328] ; Open Project of the Beijing Key Laboratory of Mobile Computing and Pervasive Device
WOS研究方向	Chemistry ; Engineering ; Instruments & Instrumentation
WOS类目	Chemistry, Analytical ; Engineering, Electrical & Electronic ; Instruments & Instrumentation
WOS记录号	WOS:000564780200001
出版者	MDPI
引用统计	被引频次：12[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/15762
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Luo, Haiyong
作者单位	1.Beijing Univ Posts & Telecommun, Sch Software Engn, Beijing 100876, Peoples R China 2.Chinese Acad Sci, Inst Comp Technol, Beijing Key Lab Mobile Comp & Pervas Device, Beijing 200190, Peoples R China
推荐引用方式 GB/T 7714	Jia, Hongwei,Luo, Haiyong,Wang, Hao,et al. ADST: Forecasting Metro Flow Using Attention-Based Deep Spatial-Temporal Networks with Multi-Task Learning[J]. SENSORS,2020,20(16):23.
APA	Jia, Hongwei.,Luo, Haiyong.,Wang, Hao.,Zhao, Fang.,Ke, Qixue.,...&Zhao, Yunyun.(2020).ADST: Forecasting Metro Flow Using Attention-Based Deep Spatial-Temporal Networks with Multi-Task Learning.SENSORS,20(16),23.
MLA	Jia, Hongwei,et al."ADST: Forecasting Metro Flow Using Attention-Based Deep Spatial-Temporal Networks with Multi-Task Learning".SENSORS 20.16(2020):23.