Abstract

As a crucial component in intelligent transportation systems, traffic flow prediction has recently attracted widespread research interest in the field of artificial intelligence (AI) with the increasing availability of massive traffic mobility data. Its key challenge lies in how to integrate diverse factors (such as temporal rules and spatial dependencies) to infer the evolution trend of traffic flow. To address this problem, we propose a unified neural network called Attentive Traffic Flow Machine (ATFM), which can effectively learn the spatial-temporal feature representations of traffic flow with an attention mechanism. In particular, our ATFM is composed of two progressive Convolutional Long Short-Term Memory (ConvLSTM [1]) units connected with a convolutional layer. Specifically, the first ConvLSTM unit takes normal traffic flow features as input and generates a hidden state at each time-step, which is further fed into the connected convolutional layer for spatial attention map inference. The second ConvLSTM unit aims at learning the dynamic spatialtemporal representations from the attentionally weighted traffic flow features. Further, we develop two deep learning frameworks based on ATFM to predict citywide short-term/long-term traffic flow by adaptively incorporating the sequential and periodic data as well as other external influences. Extensive experiments on two standard benchmarks well demonstrate the superiority of the proposed method for traffic flow prediction. Moreover, to verify the generalization of our method, we also apply the customized framework to forecast the passenger pickup/dropoff demands in traffic prediction and show its superior performance.Our code and data are available at https://github.com/liulingbo918/ATFM.

Framework

In this section, we propose a unified neural network, named Attentive Traffic Flow Machine (ATFM), to learn the spatialtemporal representations of traffic flow. ATFM is designed to adequately capture various contextual dependencies of the traffic flow, e.g., the spatial consistency and the temporal dependency of long and short term.

Experiment

Conclusion

In this work, we utilize massive human trajectory data collected from mobility digital devices to study the traffic flow prediction problem. Its key challenge lies in how to adaptively integrate various factors that affect the flow changes, such as sequential trends, periodic laws and spatial dependencies. To address these issues, we propose a novel Attentive Traffic Flow Machine (ATFM), which explicitly learns dynamic spatialtemporal representations from historical traffic flow maps with an attention mechanism. Based on the proposed ATFM, we develop a unified framework to adaptively merge the sequential and periodic representations with the aid of a temporallyvarying fusion module for citywide traffic flow prediction. By conducting extensive experiments on two public benchmarks, we have verified the effectiveness of our method for traffic flow prediction. Moreover, to verify the generalization of ATFM, we apply the customized framework to forecast the passenger pickup/dropoff demand and it can also achieve practical performance on this traffic prediction task.

However, there is still much room for improvement. First, it may be suboptimal to divide the studied cities into regular grid maps. In future work, we would divide them into traffic analysis zones with irregular shapes on the basis of the functionalities of regions. We would model such traffic systems as graphs and adapt Graph Convolutional Network (GCN [1], [2]) to learn spatial-temporal features. Second, the functionality information of zones has not been fully explored in most previous works. Intuitively, the zones with the same functionalities usually have similar traffic flow patterns. For instance, most residential regions have high outflow during morning rush hours and have high inflow during evening rush hours. Base on this consideration, we plan to incorporate the prior knowledge of functionality information of zones (e.g., the Point of Interest (POI) data, land-use data and sociodemographic data) into GCN to further improve the prediction accuracy.

References

[1] D. K. Duvenaud, D. Maclaurin, J. Iparraguirre, R. Bombarell, T. Hirzel, A. Aspuru-Guzik, and R. P. Adams, “Convolutional networks on graphs for learning molecular fingerprints,” in NIPS, 2015, pp. 2224–2232.

[2] J. Chen, L. Liu, H. Wu, J. Zhen, G. Li, and L. Lin, “Physical-virtual collaboration graph network for station-level metro ridership prediction,” arXiv preprint arXiv:2001.04889, 2020.