Spatio-temporal Correlation Research Articles

Physics-guided spatiotemporal approach for melt pool size prediction in laser powder bed fusion

ABSTRACT This study first proposes a physics-guided spatiotemporal graph. The melt pool temporal feature nodes and spatial feature nodes are extracted from melt pool images based on spatiotemporal correlations. These feature points are treated as graph nodes, and the edge weights in the graph are calculated based on an analytical heat transfer model. It supplements the physical information of accumulated heat. Based on this, the graph convolutional network (GCN) is used to extract spatiotemporal information of heat transfer, and the long short-term memory (LSTM) is used to capture the evolution patterns of the melt pool. Thus, a melt pool spatiotemporal feature extraction model is constructed for melt pool size prediction. The model achieves prediction errors within 12% for the melt pool area and 8.5% for the melt pool perimeter. Subsequently, interpretability analysis indicates that the physics-guided approach enhances the model's learning ability for the cumulative contribution of heat from neighboring tracks.

Micro- and mesoplastic pollution in the surface water and nekton from the Eastern Indian Ocean: Spatiotemporal variation, correlation and risk assessment.

The pollution of micro- and mesoplastic (MMP) in the Eastern Indian Ocean (EIO) remains poorly understood. The present study revealed that MMP abundance in nekton from EIO in 2022 (mean: 2.30 ± 0.39 items individual-1 and 1.81 ± 0.54 items g-1) was significantly higher than that in 2021 (mean: 1.60 ± 0.22 items individual-1 and 0.80 ± 0.13 items g-1). In contrast, MMP abundance in surface water varied insignificantly between 2021 (mean: 0.04 ± 0.01 items m-3) and 2022 (mean: 0.05 ± 0.02 items m-3). The rise in predominant polymers-polypropylene (PP), rayon (RA), and polyester (PES)-in nekton from 2021 to 2022 may suggest increased pollution from face masks and home textiles along coastal regions. Notable spatial variation in PP and RA between the northeastern and southeastern regions was observed only in nekton, suggesting they are better indicators of MMP spatiotemporal variation than surface water. Shadow driftfish ingested more MMPs than purpleback flying squid and mackerel scad, likely due to its deeper habitat. By simultaneously considering color, composition, and shape, integrated MMP analysis showed insignificant correlation between MMP pollution in surface water and nekton, suggesting that nekton may ingest MMPs through multiple pathways beyond surface water. Risk indices for surface water and nekton reached moderate to upper levels globally, emphasizing the need for continued monitoring in the EIO. Epoxy resin, rubber, and PP + acrylic were identified as the most hazardous polymers, providing a valuable basis for developing effective strategies to mitigate plastic pollution.

Short-term interval prediction strategy of photovoltaic power based on meteorological reconstruction with spatiotemporal correlation and multi-factor interval constraints

Short-term photovoltaic (PV) power interval prediction provides a basis for day-ahead power dispatching and generation planning. However, the current gridded numerical weather prediction (NWP) has poor matching in specific PV stations, and the lack of consideration of PV power mutation characteristics and historical correlation in interval prediction, which further limit the improvement of PV power prediction accuracy. In this regard, this paper proposes a novel short-term interval prediction strategy for PV power. Based on the second-order extended hidden Markov model (HMM), the key meteorological elements of the PV station with poor matching are reconstructed. In the interval prediction, the trend mutation and historical correlation characteristics of the PV sequence are fully considered, and a PV power interval prediction method that combines three factors such as trend change, time correlation and numerical mutation is proposed. The proposed method is applied to a PV station in Jilin, China. The results show that compared with other methods, the RMSE of the proposed method is reduced by 5.3 % on average, and the CWC is reduced by at least 2.1 %, which verifies the effectiveness of the proposed method.

A groundwater level spatiotemporal prediction model based on graph convolutional networks with a long short-term memory

ABSTRACT The performance of regional groundwater level (GWL) prediction model hinges on understanding intricate spatiotemporal correlations among monitoring wells. In this study, a graph convolutional network (GCN) with a long short-term memory (LSTM) (GCN–LSTM) model is introduced for GWL prediction utilizing data from 16 wells located in the northeastern Xiangtan City, China. This model is designed to account for both the hybrid temporal dependencies and spatial autocorrelations among wells. It consists of two parts: the spatial part employs GCNs to extract spatial characteristics from a spatial self-similarity weight matrix and an attribute self-similarity wight matrix among wells; the temporal part utilizes a LSTM module to capture the temporal patterns of GWL sequences, along with monthly precipitation and temperature data. This model dynamically predicts changes in groundwater levels, achieving higher accuracy on average compared to single-well predictions using LSTM. By incorporating both temporal dependencies and spatial autocorrelations, the GCN–LSTM model demonstrated an average improvement in goodness-of-fit of approximately 11.21% over the LSTM-based model for individual wells. Its application holds significant reference value for the sustainable utilization and development of groundwater resources in Xiangtan City.

Differences in the Effects of Three Water Elements on Vegetation in Different Climatic Regions: Insights From the Yellow River Basin, China

ABSTRACTThe effects of water on vegetation have always been a concern. It is an important support as well as a major limiting factor with respect to vegetation growth. By analysing the spatiotemporal changes and correlations between precipitation (PRE), soil moisture (SM), vapour pressure deficit (VPD) and normalized difference vegetation index (NDVI) in the Yellow River Basin, we explored the different effects of three water elements on vegetation in different climatic regions. Our findings reveal the following: (1) NDVI and the three water elements report an increasing trend in the Yellow River Basin, with NDVI increasing most significantly. 92.1% of the Yellow River Basin showed an increase in NDVI. (2) Vegetation in the Yellow River Basin was most positively affected by PRE, followed by SM and VPD. PRE mainly affected the natural vegetation on both sides of the boundary between the arid and semi‐arid regions and the semi‐humid regions. SM mainly affected the natural vegetation in the arid and semi‐arid regions, whereas VPD mainly affected the crops in the irrigation areas, and the irrigation areas in arid regions were affected the most. These findings contribute to a deeper understanding of the relationship between water elements and vegetation.

A spatiotemporal correlation and attention-based model for pipeline deformation prediction in foundation pit engineering

In foundation pit engineering, the deformation prediction of adjacent pipelines is crucial for construction safety. Existing approaches depend on constitutive models, grey correlation prediction, or traditional feedforward neural networks. Due to the complex hydrological and geological conditions, as well as the nonstationary and nonlinear characteristics of monitoring data, this problem remains a challenge. By formulating the deformation of monitoring points as multivariate time series, a deep learning-based prediction model is proposed, which utilizes the convolutional neural network to extract the spatial dependencies among various monitoring points, and leverages the bi-directional long-short memory unit network to extract temporal features. Notably, an attention mechanism is introduced to adjust the trainable weights of spatial-temporal features extracted in the prediction. The evaluation of a real-world subway project demonstrates that the proposed model has advantages compared with current models, particularly in long-term prediction. It improves the Adjusted R2 index averagely by from 19.4 to 61.6%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\%$$\\end{document} compared with existing models. The proposed model also exhibits a decrease in mean absolute error ranging from 51.5 to 70.3%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\%$$\\end{document} compared to others. Experiments and analyses verify that the spatial-temporal dependencies in time series and the attention learning for spatial-temporal features can improve the prediction of such engineering problems.

Inner–outer interaction of unsteady turbulent flow inside duct with wall corrugations: Augmentation and saturation

This study investigates the inner–outer interaction of the unsteady turbulent flow inside a duct with wall corrugations, where “inner” refers to the trapped flow in a corrugation-induced cavity array and “outer” relates to the mainstream flow transporting through a circular duct. Configurations with different pitch–diameter ratios (P/D) are used to demonstrate the effect of the cavity flow pattern on the mainstream flow variations. An improved delayed detached-eddy simulation with dynamic blending function is performed to acquire high-fidelity turbulent flow data, in which dynamic evolution of multi-scale vortex structures containing hairpin vortices and vortex fragments is clearly resolved. The subsequent statistical analysis reveals a nonlinear variation tendency of the inner–outer interaction intensity, experiencing sudden augmentation first and then a gradual attenuation toward the final saturation. Comparatively, a larger pitch ratio results in a stronger interaction under the effect of bicentric recirculation zones within the cavity array. Moreover, a proper orthogonal decomposition analysis allows for the visualization of energetic flow structures, as consistent velocity variations throughout the duct volume are identified for a larger pitch ratio, while discontinuity at the duct termination is found for a smaller pitch ratio. Finally, an advanced elliptical model based on the spatiotemporal cross correlation analysis is proposed to examine the convection velocity and sweeping velocity of the interactive mainstream flow and cavity flow. The results highlighted the presence of first-augmented and then saturated dynamics and kinematics inside the duct with the cavity array.

A day-ahead wind speed correction method: Enhancing wind speed forecasting accuracy using a strategy combining dynamic feature weighting with multi-source information and dynamic matching with improved similarity function

Forecasting error of day-ahead wind speed (WS) seriously affects wind power integration and power system security and stability. In this regard, this paper fully considers the spatiotemporal correlation of wind farms (WFs) in different geographical locations, and proposes a day-ahead WS combined correction method that integrates multi-source station dynamic information weighting. Different from the previous WS correction methods, this paper fully considers the dynamic correlation of WS between the WFs, introduces an improved weighted similarity function to screen and dynamically weight the information of WFs with dynamic correlation, and introduces the dynamic weighting feature into the WS correction process. A combined decomposition mechanism is proposed, which combines sequential variational mode decomposition (SVMD) and feature mode decomposition (FMD) models to extract the most relevant trend components and non-stationary components of forecasted and measured WS. A combined correction model is introduced, and a combined architecture of Non-stationary Transformer combined with bidirectional long short-term memory network (Ns-Transformer-BILSTM) is used to correct the stationary WS component. A dynamic matching mechanism of fluctuation components considering improved similarity is proposed for the correction of non-stationary components. The proposed method is applied to several regional WFs in China. The experimental results show that the average correction of NRMSE, NMAE and R can reach 2.4 % ∼ 3.7 %, 2.0 % ∼ 3.0 % and 3.3 % ∼ 9.7 %, respectively. The NRMSE and NMAE corresponding to the corrected WS of certain individual WFs can be reduced by 10 % and 9 %, respectively, and R can be increased by 33 %.

Spatiotemporal information propagation in confined supersonic reacting flows

The interplay between mass injection, heat release, and boundary layer development plays a key role in dictating the dynamics and stability of confined supersonic flows. The relative impacts of these factors and the timescales over which they influence the upstream and downstream flow can provide critical insights into how different operating modes develop. As such, this work presents a series of simulations of an experimental axisymmetric direct connect flowpath. The mass flow rates and chemical compositions of the injection stages are varied, and subsequent information propagation and mode transitions are analyzed using spatiotemporal correlations of cross-sectional-averaged quantities. Increasing the injection flow rate decreases the time lags and durations of positive correlations between pressure and heat release at various points along the flowpath. Meanwhile, in dual-mode cases with lower injection flow rates, these correlations develop after longer time delays and persist for a longer times, illustrating how information propagates more gradually in these scenarios. Over the full flowpath, positive correlations persist for comparatively long times between (1) the upstream isolator pressure and the pressure elsewhere, and (2) the pressure in the downstream diverging combustor section and the upstream pressure. As such, the influence of the pressure in the intermediate constant-area combustor section decays more rapidly. Conditional statistics suggest that flow blockage and pressurization from the injected mass reduce the local ignition delay, thereby facilitating increased pressurization via heat release in a positive feedback loop.

Water Supply Pipeline Operation Anomaly Mining and Spatiotemporal Correlation Study

Water Supply Pipeline Operation Anomaly Mining and Spatiotemporal Correlation Study

CTIFTrack: Continuous Temporal Information Fusion for object track

In visual tracking tasks, researchers usually focus on increasing the complexity of the model or only discretely focusing on the changes in the object itself to achieve accurate recognition and tracking of the moving object. However, they often overlook the significant contribution of video-level linear temporal information fusion and continuous spatiotemporal mapping to tracking tasks. This oversight may lead to poor tracking performance or insufficient real-time ability of the model in complex scenes. Therefore, this paper proposes a real-time tracker, namely Continuous Temporal Information Fusion Tracker (CTIFTrack). The key of CTIFTrack lies in its well-designed Temporal Information Fusion (TIF) module, which cleverly performs a linear fusion of the temporal information between the (t-1)-th and the t-th frames and completes the spatiotemporal mapping. This enables the tracker to better understand the overall spatiotemporal information and contextual spatiotemporal correlations within the video, thereby having a positive impact on the tracking task. In addition, this paper also proposes the Object Template Feature Refinement (OTFR) module, which effectively captures the global information and local details of the object, and further improves the tracker’s understanding of the object features. Extensive experiments are conducted on seven benchmarks, such as LaSOT, GOT-10K, UAV123, NFS, TrackingNet, VOT2018 and OTB-100. The experimental results validate the significant contribution of the TIF module and OTFR module to the tracking task, as well as the effectiveness of CTIFTrack. It is worth noting that while maintaining excellent tracking performance, CTIFTrack also shows outstanding real-time tracking speed. On the Nvidia Tesla T4-16GB GPU, the FPS of CTIFTrack reaches 71.98. The code and demo materials will be available at https://github.com/vpsg-research/CTIFTrack.

A multi-task spatio-temporal fusion network for offshore wind power ramp events forecasting

With the accelerated development of offshore wind farms, the impact of Wind Power Ramp Events (WPRE) on power systems has become more pronounced. Accurate prediction of WPRE is essential for mitigating their adverse effects on the grid. However, current studies on offshore WPRE are limited, especially regarding their spatio-temporal correlations and variability across large wind farm clusters. To address this, a novel forecasting approach using a multi-task spatio-temporal fusion network is presented. This method improves WPRE prediction by integrating spatial and temporal data and utilizing multitask learning to characterize ramp features. Firstly, an index system for WPRE characterization is developed, including metrics such as power change rate and duration. Based on this, an X-means classification method for WPRE is established. Secondly, a spatio-temporal encoder combining graph convolutional networks and temporal attention mechanisms has been proposed to model the spatiotemporal dependencies of meteorological changes in offshore wind farms. Finally, a multitask learning framework has been introduced to predict WPRE characteristic indices and occurrence probabilities. This approach enhances feature extraction efficiency through the joint feedback of multiple related tasks, thereby improving prediction accuracy. A case study using data from five offshore wind farms in Eastern China validates the efficacy of the proposed method.

Information-guided adaptive learning approach for active surveillance of infectious diseases

The infectious disease surveillance system is a key support tool for public health decision making. Current research concentrates on optimizing static sentinel deployment to address the problem of incomplete data due to the lack of sufficient surveillance resources. In this study, we introduce an information-guided adaptive learning strategy for the dynamic surveillance of infectious diseases. The goal is to improve monitoring effectiveness in situations where it is possible to adjust the focus of surveillance, such as serial surveys and allocation of testing tools. Specifically, we develop a probabilistic neural network model to learn spatio-temporal correlations among the numbers of infections. Based on a probabilistic model, we evaluate the information gain of monitoring a spatio-temporal target and design a greedy selection algorithm for monitoring targets selection. Moreover, we integrate two major surveillance objectives, i.e., informativeness and coverage, in the monitoring target selection. The experimental results on the synthetic dataset and two real-world datasets demonstrate the effectiveness of our approach, showcasing the promise of further exploration and application of dynamic adaptive active surveillance.

MGHCN: Multi-graph structures and hypergraph convolutional networks for traffic flow prediction

Accurate and timely traffic flow predictions are essential for effective traffic management and congestion reduction. However, most traditional prediction methods often fail to capture the complex dynamics and correlations within traffic flows due to insufficient processing of spatiotemporal data. Specifically, these methods struggle to integrate and analyze the multi-layered spatial and temporal interactions inherent in traffic data, leading to suboptimal prediction accuracy and robustness. To address this limitation, this paper presents a Multi-Graph Structures and Hypergraph Convolutional Network (MGHCN) that combines diverse graphs and hypergraphs. The MGHCN simplifies the predictive framework by integrating key components that improve its robustness and accuracy. One of the most critical components is the dual hypergraph structure, which captures edge correlations by converting traditional graph edges into hypergraph nodes. To better capture the spatiotemporal correlation of traffic data, a Graph Convolutional Network (GCN) is employed to analyze these hypergraphs in depth. Finally, a novel adjacency matrix and a dynamic graph module are used to accurately simulate interactions between spatiotemporal features, thereby enhancing the accuracy and robustness of predictions. Experimental validation on four distinct real-world traffic datasets shows that MGHCN outperforms existing state-of-the-art traffic prediction methods.

Single-shot echo planar time-resolved imaging for multi-echo functional MRI and distortion-free diffusion imaging.

To develop a single-shot SNR-efficient distortion-free multi-echo imaging technique for dynamic imaging applications. Echo planar time-resolved imaging (EPTI) was first introduced as a multi-shot technique for distortion-free multi-echo imaging. This work aims to develop single-shot EPTI (ss-EPTI) to achieve improved robustness to motion/physiological noise, increased temporal resolution, and higher SNR efficiency. A new spatiotemporal encoding that enables reduced phase-encoding blips and minimized echo spacing under the single-shot regime was developed, which improves sampling efficiency and enhances spatiotemporal correlation in the k-TE space for improved reconstruction. A continuous readout with minimized deadtime was employed to optimize SNR efficiency. Moreover, k-TE partial Fourier and simultaneous multi-slice acquisition were integrated for further acceleration. ss-EPTI provided distortion-free imaging with densely sampled multi-echo images at standard resolutions (e.g., ˜1.25 to 3 mm) in a single-shot. Improved SNR efficiency was observed in ss-EPTI due to improved motion/physiological-noise robustness and efficient continuous readout. Its ability to eliminate dynamic distortions-geometric changes across dynamics due to field changes induced by physiological variations or eddy currents-further improved the data's temporal stability. For multi-echo fMRI, ss-EPTI's multi-echo images recovered signal dropout in short- regions and provided TE-dependent functional information to distinguish non-BOLD noise for further tSNR improvement. For diffusion MRI, it achieved shortened TEs for improved SNR and provided images free from both B0-induced and diffusion-encoding-dependent eddy-current-induced distortions with multi-TE diffusion metrics. ss-EPTI provides SNR-efficient distortion-free multi-echo imaging with comparable temporal resolutions to ss-EPI, offering a new acquisition tool for dynamic imaging.

An Efficient Attentive-GRU Structure for Uncertainty Modelling of Crowdsourced Human Trajectories Under Building-obscured Urban Scenes

Abstract. Uncertainty modelling is regarded as one of the core components in the field of human mobility analysis and urban navigation, that can affect the performance of human behaviour modelling and location information acquisition. Existing uncertainty modelling algorithms towards the human movement trajectory are subjected to random and highly dynamic human motion characteristics and sampling and observation errors of Global Navigation Satellite System (GNSS) signals caused by the occlusion of buildings in urban scenes, which lead to the insufficient spatiotemporal correlation and poor accuracy of uncertainty modelling. To fill in this gap, this paper proposes an efficient attentive-GRU structure for uncertainty modelling of crowdsourced human trajectories under building-obscured urban scenes, that takes into account both temporal correlation and spatial correlation of human-originated GNSS trajectories and related motion features. A period of human motion data is modelled instead of only one or adjacent location points to avoid interference factors caused by the obstruction of urban buildings, and time-varying measurement and sampling errors are further estimated and combined with comprehensive human motion features to improve the accuracy of final uncertainty modelling. Comprehensive experiments indicate that compared with existing uncertainty modelling methods including physical models and deep-learning models, the proposed attentive-GRU structure realizes much better performance under different accuracy indexes.

An Enhanced Seamless Localization Framework Using Spatial-temporal Uncertainty Predictor Under Obscured Indoor and Outdoor Scenes

Abstract. Uncertainty modelling is regarded as one of the core components in the field of urban navigation, that can affect the performance of indoor and outdoor location information acquisition, especially under obscured scenarios. Existing multi-source fusion-based seamless positioning algorithms are subjected to random and highly dynamic human motion characteristics and changeable observation errors caused by the dynamic occlusions of human and buildings under urban scenes, which lead to the insufficient spatiotemporal correlation and poor accuracy of final multi-source fusion structure. To fill in this gap, this paper proposes an enhanced seamless localization framework using spatial-temporal uncertainty predictor under obscured indoor and outdoor scenes (ESL-STUP), that takes into account both temporal correlation and spatial correlation of trajectories provided by Wi-Fi, GNSS, and sensor-originated motion information. An iPDR-based trajectory estimation structure is proposed, using the integration of INS/PDR mechanizations, magnetic observations, and deep-learning based speed estimation to enhance the performance of traditional PDR algorithm. A period of human motion features extracted from hybrid location sources are modelled instead of only one or adjacent location points to realize time-varying measured uncertainty errors prediction, and the predicted uncertainty errors of different indoor and outdoor location sources are integrated with iPDR to realize robust seamless positioning performance. Comprehensive experiments indicate that compared with existing multi-source fusion-based seamless positioning structure, the proposed ESL-STUP realizes much better performance under different scenes.

Contaminants and residues have varied distributions in large volumes of wheat

Analysis of bulk wheat consignments for naturally-occurring contaminants and residues from plant protection products is common, and helps manage potential health risks to consumers. The heterogeneous distribution of some mycotoxins in wheat has been described, however the distribution of other contaminants and residues has not yet been reported. This study characterized distributions of deoxynivalenol, ochratoxin A, ergot alkaloids, cadmium, and glyphosate in nine large consignments of wheat by analysing composite samples representing sub-lots prepared from increments obtained during the entire loading process. The widest span of concentrations within a consignment occurred for ochratoxin A (<0.5–22.9 µg/kg) and ergot alkaloids (0.009–0.486 mg/kg), followed by deoxynivalenol (<0.05–0.76 mg/kg) and glyphosate (<0.3–5.01 mg/kg), and then cadmium (0.022–0.102 mg/kg). Experimental semivariograms were plotted to model the spatio-temporal correlation of analytes within consignments during loading. Analyses demonstrated that distributions of contaminants and residues within a particular consignment differed, and that distributions of a particular contaminant or residue differed among consignments. The results indicate that sampling during only a portion of a loading or unloading process can result in a composite sample that is not representative of the consignment and thus increase the risk of misclassifying a consignment as compliant.

Ergodic theory of diagonal orthogonal covariant quantum channels

We analyse the ergodic properties of quantum channels that are covariant with respect to diagonal orthogonal transformations. We prove that the ergodic behaviour of a channel in this class is essentially governed by a classical stochastic matrix. This allows us to exploit tools from classical ergodic theory to study quantum ergodicity of such channels. As an application of our analysis, we study dual unitary brickwork circuits which have recently been proposed as minimal models of quantum chaos in many-body systems. Upon imposing a local diagonal orthogonal invariance symmetry on these circuits, the long-term behaviour of spatio-temporal correlations between local observables in such circuits is completely determined by the ergodic properties of a channel that is covariant under diagonal orthogonal transformations. We utilize this fact to show that such symmetric dual unitary circuits exhibit a rich variety of ergodic behaviours, thus emphasizing their importance.

Dynamic Spatial-Temporal Memory Augmentation Network for Traffic Prediction.

Traffic flow prediction plays a crucial role in the development of smart cities. However, existing studies face challenges in effectively capturing spatio-temporal contexts, handling hierarchical temporal features, and understanding spatial heterogeneity. To better manage the spatio-temporal correlations inherent in traffic flow, we present a novel model called Dynamic Spatio-Temporal Memory-Augmented Network (DSTMAN). Firstly, we design three spatial-temporal embeddings to capture dynamic spatial-temporal contexts and encode the unique characteristics of time units and spatial states. Secondly, these three spatial-temporal components are integrated to form a multi-scale spatial-temporal block, which effectively extracts hierarchical spatial-temporal dependencies. Finally, we introduce a meta-memory node bank to construct an adaptive neighborhood graph, implicitly representing spatial relationships and enhancing the learning of spatial heterogeneity through a secondary memory mechanism. Evaluation on four public datasets, including METR-LA and PEMS-BAY, demonstrates that the proposed model outperforms benchmark models such as MTGNN, DCRNN, and AGCRN. On the METR-LA dataset, our model reduces the MAE by 4% compared to MTGNN, 6.9% compared to DCRNN, and 5.8% compared to AGCRN, confirming its efficacy in traffic flow prediction.