Abstract
In this paper, we study the prediction of traffic flow in the presence of missing information from data set. Specifically, we adopt three different patterns to model the missing data structure, and apply two types of approaches for the imputation. In consequence, a forecasting model via deep learning based methods is proposed to predict the traffic flow from the recovered data set. The experiments demonstrate the effectiveness of using deep learning based imputation in improving the accuracy of traffic flow prediction. Based on the experimental results, we conduct a thorough discussion on the appropriate methods to predict traffic flow under various missing data conditions, and thus shedding the light for a practical design.
Highlights
Intelligent Transportation systems (ITSs), which aims at providing innovative services and making safer, more convenient use of traffic network, typically depend on traffic flow information, i.e., the number of vehicles crossing a specific region per unit time interval, as inputs to make the underlying decision logic [1], [2]
INDEX OF PERFORMANCE In order to evaluate the effectiveness of the imputation methods, as well as the accuracy of traffic flow prediction, we introduce three different metrics, i.e., the mean absolute error (MAE), the mean relative error (MRE), and the normalized mean square error (NMSE)
The performance of mean of other lanes (MOL) imputation method in higher missing rates is better than the deep learning based imputation methods, the prediction methods with deep learning based imputation still achieve a approximative performance with the MOL imputation based traffic flow prediction method
Summary
Intelligent Transportation systems (ITSs), which aims at providing innovative services and making safer, more convenient use of traffic network, typically depend on traffic flow information, i.e., the number of vehicles crossing a specific region per unit time interval, as inputs to make the underlying decision logic [1], [2]. In the domain of ANNs based methods, deep learning is considered as one of the most effective and efficient traffic flow prediction approaches. Three deep learning methods including Stacked Autoencoders (SAEs), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) are used to extract generic traffic flow features for prediction with the filled traffic data. Different from other traffic flow prediction methods, the proposed scheme can effectively solve the traffic data missing problem occurred in the real world settings, and obtain the accurate prediction results of traffic flow. Deep learning based imputation and prediction (DLIP) model is proposed to improve the accuracy of traffic flow prediction under missing data simultaneously. For the normal scenarios with less than 40 percent data missing, deep learning based imputation methods like SAEs, LSTM and GRU can achieve less error of the prediction than the traditional ones.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
