TY - GEN
T1 - Spatiotemporal multi-Task learning for citywide passenger flow prediction
AU - Zhong, Runxing
AU - Lv, Weifeng
AU - Du, Bowen
AU - Lei, Shuo
AU - Huang, Runhe
N1 - Publisher Copyright:
© 2017 IEEE.
PY - 2018/6/26
Y1 - 2018/6/26
N2 - Massive data collected by automated fare collection (AFC) systems provides unprecedented opportunities for studying and predicting citywide passenger flow, which can be beneficial for travel planning, traffic management, and public safety. However, it is very challenging to model and predict citywide passenger flow because there are a variety of factors potentially influencing it, such as the dynamic change of passenger flow and spatiotemporal correlations. To this end, in this paper, we first investigate multiple types of heterogeneous data that are related to the citywide passenger flow, and extract different features from each type of the data, including mobility related features, connectivity and traffic capacity, regional characteristics, event and weather, and temporal view features. Then, upon these features, we develop a spatiotemporal multi-Task learning based regression approach for predicting the citywide passenger flow. Finally, we leverage real-world data sets from multiple sources for model training and validation. Experimental results on real-world data demonstrate the effectiveness of our proposed approach in predicting citywide passenger flow.
AB - Massive data collected by automated fare collection (AFC) systems provides unprecedented opportunities for studying and predicting citywide passenger flow, which can be beneficial for travel planning, traffic management, and public safety. However, it is very challenging to model and predict citywide passenger flow because there are a variety of factors potentially influencing it, such as the dynamic change of passenger flow and spatiotemporal correlations. To this end, in this paper, we first investigate multiple types of heterogeneous data that are related to the citywide passenger flow, and extract different features from each type of the data, including mobility related features, connectivity and traffic capacity, regional characteristics, event and weather, and temporal view features. Then, upon these features, we develop a spatiotemporal multi-Task learning based regression approach for predicting the citywide passenger flow. Finally, we leverage real-world data sets from multiple sources for model training and validation. Experimental results on real-world data demonstrate the effectiveness of our proposed approach in predicting citywide passenger flow.
UR - https://www.scopus.com/pages/publications/85050207839
U2 - 10.1109/UIC-ATC.2017.8397485
DO - 10.1109/UIC-ATC.2017.8397485
M3 - 会议稿件
AN - SCOPUS:85050207839
T3 - 2017 IEEE SmartWorld Ubiquitous Intelligence and Computing, Advanced and Trusted Computed, Scalable Computing and Communications, Cloud and Big Data Computing, Internet of People and Smart City Innovation, SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI 2017 - Conference Proceedings
SP - 1
EP - 8
BT - 2017 IEEE SmartWorld Ubiquitous Intelligence and Computing, Advanced and Trusted Computed, Scalable Computing and Communications, Cloud and Big Data Computing, Internet of People and Smart City Innovation, SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI 2017 - Conference Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2017 IEEE SmartWorld Ubiquitous Intelligence and Computing, Advanced and Trusted Computed, Scalable Computing and Communications, Cloud and Big Data Computing, Internet of People and Smart City Innovation, SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI 2017
Y2 - 4 April 2017 through 8 April 2017
ER -