Deep Multi-View Spatial-Temporal Network for Taxi Demand Prediction

By Huaxiu Yao, Fei Wu, Jintao Ke, Xianfeng Tang, Yitian Jia, Siyu Lu, Pinghua Gong, Jieping Ye, Zhenhui Jessie Li. AAAI Conference on Artificial Intelligence, 2018

1,258 citations recorded by the source corpus.

Taxi demand prediction is an important building block to enabling intelligent transportation systems in a smart city. An accurate prediction model can help the city pre-allocate resources to meet travel demand and to reduce empty taxis on streets which waste energy and worsen the traffic congestion. With the increasing popularity of taxi requesting services such as Uber and Didi Chuxing (in China), we are able to collect large-scale taxi demand data continuously. How to utilize such big data to improve the demand prediction is an interesting and critical real-world problem. Traditional demand prediction methods mostly rely on time series forecasting techniques, which fail to model the complex non-linear spatial and temporal relations. Recent advances in deep learning have shown superior performance on traditionally challenging tasks such as image classification by learning the complex features and correlations from large-scale data. This breakthrough has inspired researchers to explore deep learning techniques on traffic prediction problems. However, existing methods on traffic prediction have only considered spatial relation (e.g., using CNN) or temporal relation (e.g., using LSTM) independently. We propose a Deep Multi-View Spatial-Temporal Network (DMVST-Net) framework to model both spatial and temporal relations. Specifically, our proposed model consists of three views: temporal view (modeling correlations between future demand values with near time points via LSTM), spatial view (modeling local spatial correlation via local CNN), and semantic view (modeling correlations among regions sharing similar temporal patterns). Experiments on large-scale real taxi demand data demonstrate effectiveness of our approach over state-of-the-art methods.

Subfields: Building and Construction, Transportation

View Paper: https://www.semanticscholar.org/paper/839c4dd710ae7a234424aeda2f1423e0ce61bd5e

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Building and Construction

In Transportation

Citations in the corpus

GMAN: A Graph Multi-Attention Network for Traffic Prediction (2019, methodology): “To model spatial correlations, researchers apply convolutional neural networks (CNN) to capture the dependencies in Euclidean space (Zhang, Zheng, and Qi 2017; Yao et al. 2018; Yao et al. 2019).”
Ridesourcing systems: A framework and review (2019, background): “Yao et al. (2018) propose a deep multi-view spatio-temporal network (DMVST-Net) framework to model both spatial and temporal demand relations.”
PDFormer: Propagation Delay-aware Dynamic Long-range Transformer for Traffic Flow Prediction (2023, methodology): “As early solutions, convolutional neural networks (CNNs) were applied to grid-based traffic data to capture spatial dependencies, and recurrent neural networks (RNNs) were used to learn temporal dynamics (Zhang, Zheng, and Qi 2017; Yao et al. 2018).”
Deep Learning on Traffic Prediction: Methods, Analysis, and Future Directions (2021, methodology): “ion of historical time steps for prediction, it does not explicitly model the temporal sequential dependency, leading to lower performance than other comparison methods to a certain extent. DMVST-Net [61] and STDN [67] considered both spatial relation and temporal sequential relation, and used LSTM ”
STG2Seq: Spatial-temporal Graph to Sequence Model for Multi-step Passenger Demand Forecasting (2019, methodology): “These methods usually employ RNN and its variants such as Long-Short Term Memory (LSTM) [ Yao et al. , 2018 ] networks to capture temporal correlation and CNN [ LeCun et al. , 2015 ] to extract spatial relationships from the whole city [ Zhang et al. , 2016a; Zhang et al. , 2017 ] or geographically…”
Learning Dynamics and Heterogeneity of Spatial-Temporal Graph Data for Traffic Forecasting (2021, background): “Yao et al. [33] integrate CNN and LSTM to jointly capture spatial and temporal dependencies for trafﬁc prediction.”
Adaptive Graph Convolutional Recurrent Network for Traffic Forecasting (2020, background): “A part of these studies [26, 27, 7, 9] utilize CNN to capture spatial correlations among near regions based on the assumption that traffic series are generated from grid-partitioned cities [27], which does not always hold.”
A Survey of Traffic Prediction: from Spatio-Temporal Data to Intelligent Transportation (2021, methodology): “Inspired by [20, 21], we can also incorporate context data for traffic prediction, such as event data (e.g., traffic accident) and meteorological data (e.g., weathers).”
A Comprehensive Survey on Graph Neural Networks (2019, methodology): “Given historical taxi demands, location information, weather data, and event features, Yao et al. [159] incorporate LSTM, CNN and network embeddings trained by LINE [ 160] to form a joint representation for each location to predict the number of taxis demanded for a location within a time interval.”
DeepMove: Predicting Human Mobility with Attentional Recurrent Networks (2018, methodology): “Because of its powerful representation ability, RNN have been applied to many fields like click prediction [44], recommendation system [13] and mobility prediction [9, 15, 20, 23, 37, 38].”
Urban Traffic Prediction from Spatio-Temporal Data Using Deep Meta Learning (2019, methodology): “In specific, the architectures of CNNs were widely used in grid data modeling like citywide flow prediction [27] and taxi demand inference [24].”
Spatial temporal incidence dynamic graph neural networks for traffic flow forecasting (2020, methodology): “• DMVST-Net [40]: DMVST-Net is a deep multi-view spatial-temporal neural network model for taxi demand prediction.”
A Survey on Graph Neural Networks for Time Series: Forecasting, Classification, Imputation, and Anomaly Detection (2023, background): “of ride demand across different regions, enabling accurate prediction of ride-hailing needs and thereby facilitating efficient fleet management [54], [138], [214], [215], [216],”
Revisiting Spatial-Temporal Similarity: A Deep Learning Framework for Traffic Prediction (2018, methodology): “further propose a method to jointly model both spatial and temporal dependencies by integrating CNN and long shortterm memory (LSTM) (Yao et al. 2018).”
Spatiotemporal Multi-Graph Convolution Network for Ride-Hailing Demand Forecasting (2019, methodology): “In (Yao et al. 2018a), the authors further improves (Yao et al. 2018b) by modeling the periodically shift problem with the attention mechanism.”
Decoupled Dynamic Spatial-Temporal Graph Neural Network for Traffic Forecasting (2022, neutral): “Recently, deep learning-based approaches [21, 24, 43] have been proposed to capture the complex spatial-temporal correlations in traffic flow.”
Dynamic Graph Convolutional Recurrent Network for Traffic Prediction: Benchmark and Solution (2021, methodology): “It is obvious that traffic condition can be significantly affected or inferred by external factors such as weather [33, 35] and POI [7, 17], which need extra collection and preprocessing.”