PDFormer: Propagation Delay-Aware Dynamic Long-Range Transformer for Traffic Flow Prediction
نویسندگان
چکیده
As a core technology of Intelligent Transportation System, traffic flow prediction has wide range applications. The fundamental challenge in is to effectively model the complex spatial-temporal dependencies data. Spatial-temporal Graph Neural Network (GNN) models have emerged as one most promising methods solve this problem. However, GNN-based three major limitations for prediction: i) Most spatial static manner, which limits ability learn dynamic urban patterns; ii) only consider short-range information and are unable capture long-range dependencies; iii) These ignore fact that propagation conditions between locations time delay systems. To end, we propose novel Propagation Delay-aware transFormer, namely PDFormer, accurate prediction. Specifically, design self-attention module dependencies. Then, two graph masking matrices introduced highlight from short- views. Moreover, delay-aware feature transformation proposed empower PDFormer with capability explicitly modeling propagation. Extensive experimental results on six real-world public datasets show our method can not achieve state-of-the-art performance but also exhibit competitive computational efficiency. visualize learned attention map make highly interpretable.
منابع مشابه
Long-range Propagation Prediction in Shallow Water
1. Centre for Marine Science and Technology, Curtin University of Technology, G.P.O. Box U1987, Perth, Western Australia 6845. email: J.Hoffman cmst.curtin.edu.au, J.Penrose cmst.curtin.edu.au 2. Nautronix Ltd., 108 Marine Tce, Fremantle, Western Australia, 6160. email: Sandro.Ghiotto nautronix.com.au 3. Maritime Operations Division, Defence Science and Technology Organisation, Building A51, H....
متن کامل(max, +) DYNAMIC SYSTEMS FOR MODELING TRAFFIC WITH LONG-RANGE DEPENDENCE
A simple first-order recurrence in a (max, +) dynamic system is numerically investigated and shown to exhibit statistical long-range dependence, characterized by slowly decaying aggregated variances and power-law evolutions of the autocorrelation and spectrum. We propose this model as a basis for a very parsimonious modeling of some long-range dependent processes such as data traffic.
متن کاملInternodal Delay Issues in Long-Range, Adaptive Traffic Forecasts
Abstract. This paper looks at five mechanisms by which traffic-controlled intersections can interact within a longrange traffic forecast and explores the computational implications of each. These mechanisms apply to networks that can adapt to changing travel behaviors over a long period of time. One mechanism relates to user-optimal equilibrium traffic assignment, which is a prerequisite for th...
متن کاملFlow Aware Traffic Management
When the TCP and IP protocols were first proposed the emphasis was on a simple best-effort service. Later on TCP was modified in order to address the issue of congestion control, that is to make sure that the senders would not overwhelm the network by sending data too fast. Over the years these protocols have shown an excellent scalability and have allowed the Internet to grow from a network co...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Proceedings of the ... AAAI Conference on Artificial Intelligence
سال: 2023
ISSN: ['2159-5399', '2374-3468']
DOI: https://doi.org/10.1609/aaai.v37i4.25556