A GNN-Based Framework for Assessing Flood Impacts on Highway Networks: Integrating Network Structural, Functional, and Social Features

Our research examines the perceptions and intentions surrounding the use of healthcare service robots. Guided by service robot acceptance model (sRAM) and stimulus-organism-response (S-O-R) model, we explore how perceptions of functional, social, emotional, and robotic features of service robots shape their trust and use in healthcare. Our research incorporated data from 398 responses collected via an online questionnaire, which was analyzed using partial least squares structural equation modeling (PLS-SEM) through the SmartPLS software, revealing that functional (ease of use), emotional (anxiety and enjoyment), and social (social interactivity and presence) features significantly influence healthcare service robots trust and use. Contrarily, usefulness—a functional feature—had no significant role in shaping healthcare service robots trust and use. Nevertheless, trust mediated perceptions relating to anxiety, ease of use, enjoyment, social interactivity, and social presence with healthcare service robots use. Interestingly, anthropomorphism—a robotic feature—had no moderating effect while subjective norms—a non-robotic feature—only moderated the impact of social interactivity on healthcare service robots use. Conclusively, our research organizes sRAM antecedents into clear, discrete categories (functional, emotional, and social) and delivers a comprehensive, structured acceptance model. This new and novel model supports systematic theory development and comparability in healthcare service robot research while also offering critical implications for enhancing the integration and utilization of service robots within healthcare. • Functional, emotional, and social features drive robot trust and use. • Usefulness does not affect robot trust or usage. • Trust mediates the impact of anxiety, ease, and social features. • Anthropomorphism does not alter robot perception effects. • Subjective norms only modify social interactivity's influence.

With the wide use of Location-Based Social Networks (LBSNs), predicting user friendship from online social relations and offline trajectory data is of great value to improve the platform service quality and user satisfaction. Existing methods mainly focus on some hand-crafted features or graph embedding models based on the user-location bipartite graph, which cannot precisely capture the latent mobility similarity for the majority of users who have no explicit co-visit behaviors and also fail to balance the tradeoff between social features and mobility features for friendship prediction. In this regard, we propose a dual subgraph-based pairwise graph neural network (DSGNN) for friendship prediction in LBSNs, which extracts a pairwise social subgraph and a trajectory subgraph to model the social proximity and mobility similarity, respectively. Specifically, to overcome the co-visit data sparsity, we design an entropy-based random walk to construct a location graph that captures the high-level correlation between locations. Based on this, we characterize the pairwise mobility similarity from trajectory level instead of location level, which is modeled by a graph neural network (GNN) on a labeled trajectory subgraph composed of the two trajectories of the target user pair. Besides, we also utilize another GNN to extract social proximity based on social subgraph of the target user pair. Finally, we propose a gate layer to adaptively balance the fusion of the social and mobility features for friendship prediction. We conduct extensive experiments on the real-world datasets and demonstrate the superiority of our approach, which outperforms other state-of-the-art methods. In particular, the comparative experiments on the trajectory level mobility similarity further validate the effectiveness of the designed trajectory subgraph-based method, which can extract predictive mobility features.

In pedestrian trajectory prediction, the prediction accuracy depends largely on the consideration of the impact of social relations on the prediction object. Social pooling and graph neural networks (GNN) are two traditional social feature processing methods, they process sparse and nonuniform social features into more intensive and uniform information. In this paper, the Social Transformer Network (STNet) was proposed based on the GNN, which is a graph attention network. After a conditional variational auto-encoder (CVAE)-based preprocessing network provided a destination prediction, a transformer network was used to process the social feature data of the past trajectory and destination information. The transformer network was based on the self-attention mechanism, and it can assign different attention weights to different social features so that more attention is paid to the social relations with greater impacts on the pedestrian's trajectory. In this paper, STNet was tested on the ETH/UCY datasets. The results showed that average displacement error (ADE) was reduced by 17.2% and final displacement error (FDE) was reduced by 14.6%, indicating that the STNet improved the prediction accuracy.

The popularity of the wireless network and the embedded system has promoted the widespread application of the Internet of Things (IoT) monitoring system which detects anomalies for discovering emergency events to avoid losses by deploying various sensors. In the face of anomaly detection algorithms for multi-dimensional data in the IoT monitoring system, the currently popular method is to implement a model based on the graph neural network (GNN). Graph attention network (GAT), a mutant of GNN, is widely used because of the attention mechanism in capturing multi-dimensional nonlinear features between nodes. However, the GAT’s weakness in processing complex temporal features leads to the fact that anomalies with distinctive temporal features are hard to detect. In this paper, we propose Time-Series Graph Attention Network (TSGAT), an innovative GAT-based model for anomaly detection. Given the simplified attention mechanism of the GAT and its weakness in extracting temporal pattern features, we embed the time-series layer to enhance the ability to process time-series data. In addition, we designed the loss function and anomaly scoring mechanism based on data reconstruction features, multidimensional features, and temporal features. TSGAT achieves the improvement of 7%-8% F1 score on two classical public datasets for IoT monitoring: SWaT and WADI, compared with the state-of-the-art GNN-based anomaly detection model.

Traffic flow prediction remains a cornerstone of intelligent transportation systems (ITS), facilitating congestion mitigation, route optimization, and sustainable urban planning. Graph Neural Networks (GNNs) have revolutionized this domain by adeptly modeling the intricate graph-structured nature of traffic networks, where nodes represent sensors or intersections and edges denote spatial relationships. Recent years (2023–2025) have witnessed a surge in scientific innovation, with several novel approaches pushing the boundaries of traffic prediction accuracy and robustness. Notably, hybrid GNN-Transformer architectures have emerged, leveraging the spatial reasoning of GNNs and the temporal sequence modeling power of Transformers to capture long-range dependencies and complex spatiotemporal patterns. Physics-informed GNNs integrate domain knowledge, such as conservation laws and traffic flow theory, directly into the learning process, enhancing interpretability and generalization to unseen scenarios. Uncertainty-aware frameworks, including Bayesian GNNs and ensemble methods, provide probabilistic forecasts, crucial for risk-sensitive applications and adaptive traffic management in volatile urban environments. This article provides a comprehensive guide to implementing GNNs for traffic flow prediction, detailing best practices in data preparation (e.g., graph construction, feature engineering, handling missing data), model training (e.g., loss functions, regularization, hyperparameter tuning), and real-time deployment (e.g., edge computing, latency optimization). We critically compare GNNs to traditional statistical and deep learning methods, highlighting their superior ability to capture non-Euclidean spatial dependencies, adapt to dynamic and evolving network topologies, and seamlessly integrate multi-modal data sources such as weather, events, and sensor readings. Empirical evidence from widely used benchmarks, including PeMS and METR-LA, demonstrates that state-of-the-art GNN models achieve up to 15–20 % improvements in accuracy metrics such as Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) over conventional baselines. These gains are attributed to the models’ capacity for dynamic graph learning, attention-based feature selection, and robust handling of heterogeneous data. Drawing on these recent innovations, this synthesis highlights GNNs' pivotal role in fostering resilient, AI-driven traffic systems for future smart cities, setting the stage for next-generation ITS solutions that are adaptive, interpretable, and scalable. In addition to these advancements, the integration of real-time sensor data and external information sources has further improved the responsiveness of traffic prediction models. Modern GNN frameworks are capable of handling large-scale urban networks, making them suitable for deployment in metropolitan areas with complex road infrastructures. The use of transfer learning and domain adaptation techniques allows models trained in one city to be effectively applied to others, reducing the need for extensive retraining. Furthermore, explainable AI approaches within GNNs are gaining traction, enabling stakeholders to understand and trust model decisions in critical traffic management scenarios. Recent research also explores the fusion of GNNs with reinforcement learning, enabling adaptive control strategies for traffic signals and congestion pricing. The scalability of GNNs ensures that they can process data from thousands of sensors in real time, supporting city-wide traffic optimization. Advances in hardware acceleration, such as GPU and edge computing, have made it feasible to deploy these models in latency-sensitive environments. Collaborative efforts between academia, industry, and government agencies are driving the adoption of GNN-based solutions in smart city initiatives. As urban mobility continues to evolve, the ability of GNNs to incorporate emerging data modalities, such as connected vehicle telemetry and mobile device traces, will be crucial for future developments. The ongoing refinement of model architectures and training protocols promises even greater accuracy and robustness in traffic flow prediction. Ultimately, the convergence of GNNs with other AI technologies is set to transform intelligent transportation systems, paving the way for safer, more efficient, and sustainable urban mobility.

Multi-dimensional shared representation learning with graph fusion network for Session-based Recommendation

FTG-Net-E: A hierarchical ensemble graph neural network for DDoS attack detection

With the development of cities, intercity highway plays a vital role in people’s daily travel. The traffic flow on the highway network is also increasingly concerned by road managers and participants. However, due to the influence of highway network topology and extra feature such as weather, accurate traffic flow prediction becomes hard to achieve. It is difficult to construct a multidimensional feature matrix and predict the traffic flow of the network at one time. A novel prediction method based on a hybrid deep learning model is proposed, which can learn multidimensional feature and predict network-wise traffic flow efficiently. The experiment shows that the prediction accuracy of this method is significantly better than existing methods, and it has a good performance during the prediction.

The frequency of natural disasters occurrences has increased and is becoming noticed by many due to the impact on economy, society, and the environment. Decision making during disasters contributes towards community safety and resilience. Decisions taken to protect people from disasters have an impact on society, economy, environment, travel patterns, and reliability and performance of transport networks. Decisions often affect the performance of transport networks and critical infrastructure systems during disasters and during demand times. Decision making and decision styles which were observed during disaster events will show the areas that the decisions have worked well or created more risks to the society. Risks to the community due to environmental disasters are high, and they are worsened when integrated with poor decision making and actions. Good decision making provides for community resilience, good connectivity between cities, improved efficiency, safety to communities, and improved network reliability to all road users during disasters. This paper will examine decision-making scenarios used during disasters and how they impacted the community and provided for community resilience. The case study will identify ways to integrate decision making into disaster risk reduction and shows the decisions made during extreme events and how they impacted on the community and transport infrastructure.

Aim/Purpose: Accurate detection of vacant parking spaces is crucial for autonomous parking. Deep learning, particularly Graph Neural Networks (GNNs), holds promise for addressing the challenges of diverse parking lot appearances and complex visual environments. Our GNN-based approach leverages the spatial layout of detected marking points in around-view images to learn robust feature representations that are resilient to occlusions and lighting variations. We demonstrate significant accuracy improvements on benchmark datasets compared to existing methods, showcasing the effectiveness of our GNN-based solution. Further research is needed to explore the scalability and generalizability of this approach in real-world scenarios and to consider the potential ethical implications of autonomous parking technologies. Background: GNNs offer a number of advantages over traditional parking spot detection methods. Unlike methods that treat objects as discrete entities, GNNs may leverage the inherent connections among parking markers (lines, dots) inside an image. This ability to exploit spatial connections leads to more accurate parking space detection, even in challenging scenarios with shifting illumination. Real-time applications are another area where GNNs exhibit promise, which is critical for autonomous vehicles. Their ability to intuitively understand linkages across marking sites may further simplify the process compared to traditional deep-learning approaches that need complex feature development. Furthermore, the proposed GNN model streamlines parking space recognition by potentially combining slot inference and marking point recognition in a single step. All things considered, GNNs present a viable method for obtaining stronger and more precise parking slot recognition, opening the door for autonomous car self-parking technology developments. Methodology: The proposed research introduces a novel, end-to-end trainable method for parking slot detection using bird’s-eye images and GNNs. The approach involves a two-stage process. First, a marking-point detector network is employed to identify potential parking markers, extracting features such as confidence scores and positions. After refining these detections, a marking-point encoder network extracts and embeds location and appearance information. The enhanced data is then loaded into a fully linked network, with each node representing a marker. An attentional GNN is then utilized to leverage the spatial relationships between neighbors, allowing for selective information aggregation and capturing intricate interactions. Finally, a dedicated entrance line discriminator network, trained on GNN outputs, classifies pairs of markers as potential entry lines based on learned node attributes. This multi-stage approach, evaluated on benchmark datasets, aims to achieve robust and accurate parking slot detection even in diverse and challenging environments. Contribution: The present study makes a significant contribution to the parking slot detection domain by introducing an attentional GNN-based approach that capitalizes on the spatial relationships between marking points for enhanced robustness. Additionally, the paper offers a fully trainable end-to-end model that eliminates the need for manual post-processing, thereby streamlining the process. Furthermore, the study reduces training costs by dispensing with the need for detailed annotations of marking point properties, thereby making it more accessible and cost-effective. Findings: The goal of this research is to present a unique approach to parking space recognition using GNNs and bird’s-eye photos. The study’s findings demonstrated significant improvements over earlier algorithms, with accuracy on par with the state-of-the-art DMPR-PS method. Moreover, the suggested method provides a fully trainable solution with less reliance on manually specified rules and more economical training needs. One crucial component of this approach is the GNN’s performance. By making use of the spatial correlations between marking locations, the GNN delivers greater accuracy and recall than a completely linked baseline. The GNN successfully learns discriminative features by separating paired marking points (creating parking spots) from unpaired ones, according to further analysis using cosine similarity. There are restrictions, though, especially where there are unclear markings. Successful parking slot identification in various circumstances proves the recommended method’s usefulness, with occasional failures in poor visibility conditions. Future work addresses these limitations and explores adapting the model to different image formats (e.g., side-view) and scenarios without relying on prior entry line information. An ablation study is conducted to investigate the impact of different backbone architectures on image feature extraction. The results reveal that VGG16 is optimal for balancing accuracy and real-time processing requirements. Recommendations for Practitioners: Developers of parking systems are encouraged to incorporate GNN-based techniques into their autonomous parking systems, as these methods exhibit enhanced accuracy and robustness when handling a wide range of parking scenarios. Furthermore, attention mechanisms within deep learning models can provide significant advantages for tasks that involve spatial relationships and contextual information in other vision-based applications. Recommendation for Researchers: Further research is necessary to assess the effectiveness of GNN-based methods in real-world situations. To obtain accurate results, it is important to employ large-scale datasets that include diverse lighting conditions, parking layouts, and vehicle types. Incorporating semantic information such as parking signs and lane markings into GNN models can enhance their ability to interpret and understand context. Moreover, it is crucial to address ethical concerns, including privacy, potential biases, and responsible deployment, in the development of autonomous parking technologies. Impact on Society: Optimized utilization of parking spaces can help cities manage parking resources efficiently, thereby reducing traffic congestion and fuel consumption. Automating parking processes can also enhance accessibility and provide safer and more convenient parking experiences, especially for individuals with disabilities. The development of dependable parking capabilities for autonomous vehicles can also contribute to smoother traffic flow, potentially reducing accidents and positively impacting society. Future Research: Developing and optimizing graph neural network-based models for real-time deployment in autonomous vehicles with limited resources is a critical objective. Investigating the integration of GNNs with other deep learning techniques for multi-modal parking slot detection, radar, and other sensors is essential for enhancing the understanding of the environment. Lastly, it is crucial to develop explainable AI methods to elucidate the decision-making processes of GNN models in parking slot detection, ensuring fairness, transparency, and responsible utilization of this technology.

Node injection attack on Graph Neural Networks (GNNs) is an emerging and practical attack scenario that the attacker injects malicious nodes rather than modifying original nodes or edges to affect the performance of GNNs. However, existing node injection attacks ignore extremely limited scenarios, namely the injected nodes might be excessive such that they may be perceptible to the target GNN. In this paper, we focus on an extremely limited scenario of single node injection evasion attack, i.e., the attacker is only allowed to inject one single node during the test phase to hurt GNN's performance. The discreteness of network structure and the coupling effect between network structure and node features bring great challenges to this extremely limited scenario. We first propose an optimization-based method to explore the performance upper bound of single node injection evasion attack. Experimental results show that 100%, 98.60%, and 94.98% nodes on three public datasets are successfully attacked even when only injecting one node with one edge, confirming the feasibility of single node injection evasion attack. However, such an optimization-based method needs to be re-optimized for each attack, which is computationally unbearable. To solve the dilemma, we further propose a Generalizable Node Injection Attack model, namely G-NIA, to improve the attack efficiency while ensuring the attack performance. Experiments are conducted across three well-known GNNs. Our proposed G-NIA significantly outperforms state-of-the-art baselines and is 500 times faster than the optimization-based method when inferring.

Graph neural networks (GNNs) have excelled in predicting graph properties in various applications ranging from identifying trends in social networks to drug discovery and malware detection. With the abundance of new architectures and increased complexity, GNNs are becoming highly specialized when tested on a few well-known datasets. However, how the performance of GNNs depends on the topological and features properties of graphs is still an open question. In this work, we introduce a comprehensive benchmarking framework for graph machine learning, called HypBench, focusing on the performance of GNNs across varied network structures. Utilizing the geometric soft configuration model in hyperbolic space, we generate synthetic networks with realistic topological properties and node feature vectors. This approach enables us to assess the impact of network properties, such as topology-feature correlation, degree distributions, local density of triangles, and homophily, on the effectiveness of different GNN architectures. Our results highlight the dependency of model performance on the interplay between network structure and node features, providing insights for model selection in various scenarios. This study contributes to the field by offering a versatile tool for evaluating GNNs, thereby assisting in developing and selecting suitable models based on specific data characteristics.

Using the Network Theory, comparisons were made with node degree distributions, betweennesses, shortest path lengths, and clustering coefficients of the provincial Expressway network and Trunk Highway network consisting partly of the Expressway System and the National Highway System in the same province. Based on structural characteristics of networks, this paper focused on comparing vulnerabilities between the provincial Expressway and Trunk Highway network. In addition, destructive experiments focusing on edge betweenness were conducted, while reactions and reliabilities from the Expressway Network and Trunk Highway Network were observed. Results showed that when 15% of all segments were damaged or blocked, the general efficiency of the Expressway System reduced to half of normal, which meant the whole system almost stopped running. On the other hand, this situation would occur when 25% of segments were damaged or blocked in the Trunk Highway Network, and this indicated the Trunk Highway Network had higher reliability than the Expressway System.

Traffic flow data are typical spatial-temporal data. Most of the traditional traffic flow prediction methods only consider the temporal correlation of data, but ignore the spatial correlation, so that the accuracy of prediction reduce. We summarizes the traffic flow prediction model based on graph neural network (GNN), starting from the basic definition of graph convolution, covering frequency domain and spatial domain graph convolution, and introducing the chebyshev network, graph convolution neural network (GCN) and graph attention network (GAT). We consider the spatial correlation of traffic flow, building three forecasting model based on Graph Neural Network, compared and analyzed the prediction results, which proves the effectiveness of the application of graph neural network in traffic flow prediction.

Edge features contain important information about graphs. However, current state-of-the-art neural network models designed for graph learning, e.g. graph convolutional networks (GCN) and graph attention networks (GAT), adequately utilize edge features, especially multi-dimensional edge features. In this paper, we build a new framework for a family of new graph neural network models that can more sufficiently exploit edge features, including those of undirected or multi-dimensional edges. The proposed framework can consolidate current graph neural network models; e.g. graph convolutional networks (GCN) and graph attention networks (GAT). The proposed framework and new models have the following novelties: First, we propose to use doubly stochastic normalization of graph edge features instead of the commonly used row or symmetric normalization approches used in current graph neural networks. Second, we construct new formulas for the operations in each individual layer so that they can handle multi-dimensional edge features. Third, for the proposed new framework, edge features are adaptive across network layers. As a result, our proposed new framework and new models can exploit a rich source of graph information. We apply our new models to graph node classification on several citation networks, whole graph classification, and regression on several molecular datasets. Compared with the current state-of-the-art methods, i.e. GCNs and GAT, our models obtain better performance, which testify to the importance of exploiting edge features in graph neural networks.