| Abstrakt: |
Monitoring traffic conditions on urban signalized road networks is an essential component of urban traffic control systems. Due to the sparseness of trajectory data and the influence of signal timing, it is challenging to estimate the traffic condition of large-scale urban signalized networks based on trajectory data. In this study, a novel and integrated data-driven learning approach (NEI-SE) is proposed, incorporating road network segmentation, speed matching, and sparse data imputation for the estimation of travel speed. First, the urban traffic network is divided according to signalized intersection and road segment length, considering the influence of signal timing on urban traffic speed. Then, based on taxi trajectory data and the divided road network, a traffic condition matrix is constructed describing the road conditions. Finally, a lightweight multi-view learning method that integrates temporal patterns and spatial topological relations is proposed to fill the missing values of the traffic condition matrix. The approach was validated on real-world traffic trajectory data collected in Wuhan, China. The results showed that NEI-SE outperformed nine existing baselines in terms of imputation accuracy. In addition, the AutoNavi congestion data was used to evaluate the data quality of the estimated traffic speed data due to lack ground truth of traffic speed. The results showed that the congestion index data had a significant negative correlation with imputed traffic speed series, with an average correlation coefficient of − 0.67, proving that the traffic speed data estimated by the proposed approach have satisfactory quality. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink