A Graph-Based Temporal Attention Framework for Multi-Sensor Traffic Flow Forecasting

Zhang, Shaokun; Guo, Yao; Zhao, Peize; Zheng, Chuanpan; Chen, Xiangqun

doi:10.1109/tits.2021.3072118

Cited by 46 publications

(21 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One approach [30] used GAT to forecast traffic metrics of the edge attributes rather than the sensor attributes, while another [14] used an attention-based method on GPS traces to get a fine-grained representation. More closely related to our present work, other works [1,60] used attention to capture temporal dynamics. Nevertheless, none of these works took into account the node and edges unique dynamics, nor provided a method to effectively capture these dynamics.…”

Section: Related Workmentioning

confidence: 67%

Because Every Sensor Is Unique, so Is Every Pair: Handling Dynamicity in Traffic Forecasting

Prabowo,

Shao,

Xue

et al. 2023

Preprint

View full text Add to dashboard Cite

Locations of the sensors on the Californian highway network surrounding the bay area. Installing a network of sensors on a road infrastructure enables traffic forecasting and smarter cities. (b) At each sensor, traffic forecasting uses the recent sensor readings (solid blue line) to predict the future traffic (red line).This forecast is made by our proposed architecture Graph Selfattention WaveNet (G-SWaN). Our forecasts accurately predict the future traffic (dotted blue line).

show abstract

Section: Related Workmentioning

confidence: 67%

Because Every Sensor Is Unique, so Is Every Pair: Handling Dynamicity in Traffic Forecasting

Prabowo,

Shao,

Xue

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Each sensor of the machine is considered as a node of the graph, and different nodes are connected to each other. As for the weight calculation of each edge of the graph, the Euclidean distance is a commonly used approach in traffic flow forecasting or the action recognition field [32,33]. However, the relationship between different sensors is implicit: the closest distance between two sensors in the Euclidean domain may not mean they are closely related.…”

Section: Hierarchical Graph Constructionmentioning

confidence: 99%

Hierarchical graph neural network with adaptive cross-graph fusion for remaining useful life prediction

Wang¹,

Zhang²,

Lu³

et al. 2023

Meas. Sci. Technol.

View full text Add to dashboard Cite

Multi-sensor monitoring data can be used to predict the remaining useful life of the complex machine system, avoid uncertainty break down, and ensure the reliability of industry process. However, previous prognostics methods focus more on the sequential data obtained from multi-sensors, fail to explicitly explore the underlying structural information of the equipment which is given in advance. To fully leverage the structural information into modeling process, and thus improve the RUL prediction performance, a Hierarchical Graph Neural Network with Adaptive Cross-Graph Fusion for remaining useful life prediction, i.e., HGNN-ACGF, is proposed in this study. In HGNN-ACGF, a hierarchical graph consists of sensor graph and module graph, is constructed by introducing the structural information to fully model the degradation trend information of the complex machine system first. Secondly, the GNN is adopted to learn the representation at both module graph and sensor graph, and an adaptive cross-graph fusion block is proposed. Owing to the cross-graph fusion block, the representation from different graphs can be fused adaptively by considering the relative importance between different modules and sensors. Finally, the remaining useful life is predicted by utilizing several fully connected layers with the graph representation. To verify the proposed method, the experiments were conducted on a set of degradation datasets of aircraft engines provided by the C-MAPSS. The experimental results show that the proposed method has superior performance in remaining useful life prediction over the state-of-the-art methods.

show abstract

“…Peixiao Wang et al [ 15 ] (2022) proposed a multi-view bidirectional spatio-temporal network based on the spatio-temporal network. Shaokun Zhang et al [ 16 ] (2022) proposed a graph-based multi-sensor prediction framework which improved the accuracy of prediction…”

Section: Literature Reviewmentioning

confidence: 99%

Multi-Section Traffic Flow Prediction Based on MLR-LSTM Neural Network

Shi

2022

Sensors

View full text Add to dashboard Cite

As the aggravation of road congestion leads to frequent traffic crashes, it is necessary to relieve traffic pressure through traffic flow prediction. As well, the traffic flow of the target road section to be predicted is also closely related to the adjacent road sections. Therefore, in this paper, a prediction method based on the combination of multiple linear regression and Long-Short-Term Memory (MLR-LSTM) is proposed, which uses the incomplete traffic flow data in the past period of time of the target prediction section and the continuous and complete traffic flow data in the past period of time of each adjacent section to jointly predict the traffic flow changes of the target section in a short time. The accurate prediction of future traffic flow changes can be solved based on the model supposed when the traffic flow data of the target road section is partially missing in the past period of time. The accuracy of the prediction results is the same as that of the current mainstream prediction results based on continuous and non-missing target link flow data. Meanwhile, there is a small-scale improvement when the data time interval is short enough. In the case of frequent maintenance of cameras in actual traffic sections, the proposed prediction method is more feasible and can be widely used.

show abstract

A Graph-Based Temporal Attention Framework for Multi-Sensor Traffic Flow Forecasting

Cited by 46 publications

References 55 publications

Because Every Sensor Is Unique, so Is Every Pair: Handling Dynamicity in Traffic Forecasting

Because Every Sensor Is Unique, so Is Every Pair: Handling Dynamicity in Traffic Forecasting

Hierarchical graph neural network with adaptive cross-graph fusion for remaining useful life prediction

Multi-Section Traffic Flow Prediction Based on MLR-LSTM Neural Network

Contact Info

Product

Resources

About