Enhancing financial fraud detection with graph neural network and ensemble learning: insights from Related Party Transactions network

Hypergraph, an expressive structure with flexibility to model the higher-order correlations among entities, has recently attracted increasing attention from various research domains. Despite the success of Graph Neural Networks (GNNs) for graph representation learning, how to adapt the powerful GNN-variants directly into hypergraphs remains a challenging problem. In this paper, we propose UniGNN, a unified framework for interpreting the message passing process in graph and hypergraph neural networks, which can generalize general GNN models into hypergraphs. In this framework, meticulously-designed architectures aiming to deepen GNNs can also be incorporated into hypergraphs with the least effort. Extensive experiments have been conducted to demonstrate the effectiveness of UniGNN on multiple real-world datasets, which outperform the state-of-the-art approaches with a large margin. Especially for the DBLP dataset, we increase the accuracy from 77.4% to 88.8% in the semi-supervised hypernode classification task. We further prove that the proposed message-passing based UniGNN models are at most as powerful as the 1-dimensional Generalized Weisfeiler-Leman (1-GWL) algorithm in terms of distinguishing non-isomorphic hypergraphs. Our code is available at https://github.com/OneForward/UniGNN.

18 - A Lagrangian framework for learning in graph neural networks

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

Graph representation learning distills the complex structures of graphs into tractable, low-dimensional vector spaces, capturing essential topological and attribute-based properties. Graph Neural Networks (GNNs) have become a pivotal tool in this domain, leveraging graph structures to iteratively update node representations through neighbor aggregations. These representations support fundamental tasks such as node classification, link prediction, and graph classification, applicable across diverse fields from social networks and biological systems to citation networks. Despite their success, GNNs face critical challenges: they often underperform on heterophilic graph data where connected nodes display dissimilar characteristics, suffer from oversmoothing which impairs performance as network depth increases, and are sensitive to hierarchical structures. Furthermore, they are vulnerable to adversarial attacks that can severely compromise model integrity. This thesis introduces the use of neural differential equations in GNNs to enhance representation learning and robustness, addressing these challenges comprehensively. The adoption of Graph Neural Differential Equation Networks (GDENs) employs a dynamic systems approach to evolve node features over continuous time, thereby enhancing the capacity of GNNs to process and learn from graph-structured data. This method governs node feature propagation through differential equations, enabling more refined control over the learning process compared to conventional methods. The initial contribution of this thesis enhances representation learning on heterophilic graphs through a neural convection-diffusion differential equation. Subsequently, the thesis explores the relationship between stability in dynamical systems and robustness within GDENs. A neural Hamiltonian differential equation model is developed, establishing energy-conservative systems within GNNs to bolster robustness against adversarial attacks. Extending beyond traditional integer-order differential equations, the thesis incorporates fractional calculus through the Fractional-Order Graph Neural Differential Equation Networks (F-GDENs) framework. This approach introduces memory and non-local interactions, boosting the networks' ability to handle hierarchical structures and mitigate oversmoothing. F-GDENs not only integrate seamlessly with existing GDENs to enhance representation learning across various datasets, but also demonstrate tighter output perturbation bounds in scenarios involving input and topology perturbations. Empirical results further validate the superior robustness of F-GDENs models compared to integer-order GDENs. In summary, this thesis advances the robustness and capacity of representation learning through GDENs by innovating with new differential equations and extending to fractional-order derivatives. These advancements establish a solid foundation for future research into robust and adaptive GNN architectures, presenting promising implications for practical applications.

Prediction of normal boiling point and critical temperature of refrigerants by graph neural network and transfer learning

Graphs are ubiquitous data structures in various fields, such as social media, transportation, linguistics and chemistry. To solve downstream graph-related tasks, it is of great significance to learn effective representations for graphs. My research strives to help meet this demand; due to the huge success of deep learning methods, especially graph neural networks, in graph-related problems, my emphasis has primarily been on improving their power for graph representation learning. More specifically, my research spans across the following three main areas: (1) robustness of graph neural networks, where we seek to study the performance of them under random noise and carefully-crafted attacks; (2) self-supervised learning in graph neural networks, where we aim to alleviate their need for costly annotated data by constructing self-supervision to help them fully exploit unlabeled data; and (3) applications of graph neural networks, where my work is to apply graph neural networks in various applications such as traffic flow prediction. This research statement, 'Graph Mining with Graph Neural Networks', is focused on my research endeavors specifically related to the aforementioned three topics.

Graph Neural Networks (GNNs) demonstrate their significance by effectively modeling complex interrelationships within graph-structured data. To enhance the credibility and robustness of GNNs, it becomes exceptionally crucial to bolster their ability to capture causal relationships. However, despite recent advancements that have indeed strengthened GNNs from a causal learning perspective, conducting an in-depth analysis specifically targeting the causal modeling prowess of GNNs remains an unresolved issue. In order to comprehensively analyze various GNN models from a causal learning perspective, we constructed an artificially synthesized dataset with known and controllable causal relationships between data and labels. The rationality of the generated data is further ensured through theoretical foundations. Drawing insights from analyses conducted using our dataset, we introduce a lightweight and highly adaptable GNN module designed to strengthen GNNs' causal learning capabilities across a diverse range of tasks. Through a series of experiments conducted on both synthetic datasets and other real-world datasets, we empirically validate the effectiveness of the proposed module. The codes are available at https://github.com/yaoyao-yaoyao-cell/CRCG.

Active learning for graph neural networks (GNNs) aims to select B nodes to label for the best possible GNN performance. Carefully selected labeled nodes can help improve GNN performance and hence motivates a line of research works. Unfortunately, existing methods still provide inferior GNN performance or cannot scale to large networks.Motivated by these limitations, in this paper, we present FICOM, an effective and scalable GNN active learning framework. Firstly, we formulate the node selection as an optimization problem where we consider the importance of a node from (i) the importance of a node during the feature propagation with a connection to the personalized PageRank (PPR), and (ii) the diversity of a node brings in the embedding space generated by feature propagation. We show that the defined problem is submodular, and a greedy solution can provide a (1-1/e)\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$(1-1/e)$$\\end{document}-approximate solution.However, a standard greedy solution requires getting the node with the maximum marginal gain of the objective score in each iteration, which incurs a prohibitive running cost and cannot scale to large datasets. As our main contribution, we present FICOM, an efficient and scalable solution that provides (1-1/e)\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$(1-1/e)$$\\end{document}-approximation guarantee and scales to graphs with millions of nodes on a single machine. The main idea is that we adaptively maintain the lower- and upper-bound of the marginal gain for each node v. In each iteration, we can first derive a small subset of candidate nodes and then compute the exact score for this subset of candidate nodes so that we can find the node with the maximum marginal gain efficiently. Extensive experiments on six benchmark datasets using four GNNs, including GCN, SGC, APPNP, and GCNII, show that our FICOM consistently outperforms existing active learning approaches on semi-supervised node classification tasks using different GNNs. Moreover, our solution can finish within 5 h on a million-node graph.

Session-based recommendation aims to predict the next clicked item using the current ongoing session of an anonymous user. Since the user's information is unknown, the available information is limited. In recent years, due to the excellent performance of graph neural networks in many applications, many works have applied graph neural networks to session-based recommendation. However, we found that the conversion of session data into graph-structured data is a lossy graph encoding method, which leads to the loss of item order information in the session. In addition, only one session can be used for recommendation in session-based recommendation. Compared with other recommendations, the problem of data sparsity is more serious. Self-supervised learning can discover ground-truth samples and has great potential in solving the data sparsity problem. Therefore, we propose a session-based recommendation model based on graph neural network and contrastive learning to solve the problem of information loss and data sparsity in graph encoding. Extensive experiments on three benchmark datasets demonstrate that our model outperforms other methods.

Accurate prediction of the octanol/water distribution coefficient (logD) is crucial for evaluating the lipophilicity, ADMET profiles, and potential of drug candidates. With the increasing interest in peptide-based therapeutics, there is a growing demand for reliable logD prediction methods specifically tailored to short peptides (2-6 residues). In this study, traditional machine learning and graph neural network (GNN) were employed to predict logD for short peptides at near-physiological pH levels (7.0-7.4). We took two dedicated peptide logD data sets and developed machine learning models using molecular fingerprints and descriptors. We also implemented five advanced GNN architectures. The WLN model achieved the best predictive performance in GNNs, with R2 of 0.907, RMSE of 0.377, and MAE of 0.278, and outperformed the best traditional model, SVM with RFECV-processed descriptors, R2 of 0.894, RMSE of 0.403, and MAE of 0.276). Additionally, we selected the first two models from each of these two types of models and constructed consensus models based on RMSE. Among them, the consensus model composed of SVM with RFECV-processed descriptors, WLN, and WeaveGNN performed the best. Its prediction accuracy (R2 of 0.923, RMSE of 0.343, MAE of 0.241) has been significantly improved compared with the basic model that constitutes it. Then, we compared this consensus model with previous studies. Although we did not achieve better results, the highlight of our research lies in its interpretability. Our further analysis identified key molecular features influencing peptide logD, including MolLogP, fr_amide, and MaxEStateIndex descriptors, as well as peptide backbone and polar functional groups. This study is the first to apply the GNN method to the logD prediction task of short peptides and systematically compares the performance of different features and algorithms in this task. At the same time, we provide an interpretable logD prediction tool for short peptides to support the design of lipophilic peptide therapies. All source code we used are freely available at https://github.com/1756403aaa-sketch/logD.

Integrating renewables and grid-edge technologies has increased modern power systems’ complexity and data intensity, necessitating advanced Artificial Intelligence (AI) for stability management. However, centralized AI faced fundamental limitations due to data privacy, sovereignty, and communication constraints. This review explored the fusion of Federated Learning (FL) and Graph Neural Networks (GNNs) as a transformative privacy-preserving paradigm. FL enabled collaborative training across decentralized utilities without raw data sharing, while GNNs natively modelled grid topology. The study systematically analyzed FL–GNN applications in state estimation, stability assessment, anomaly detection, and resilient control. Key challenges—including data heterogeneity, communication efficiency, and privacy–accuracy trade-offs—were critically examined. The review concluded by outlining future pathways, such as physics-informed models and digital twin integration, highlighting the potential of FL and GNNs for secure and intelligent grid management. Reported implementations show up to 8% higher predictive accuracy and over 10% improvement in real-time control efficiency compared to centralized models.

Complex systems are often modelled using graphs, and one of the key tasks in complex system analysis is identifying anomalies in a graph. A graph anomaly is a pattern that does not follow the typical patterns predicted by the graph's structures and/or properties. The present article provides a comprehensive review of the techniques, challenges, and advancements in the field of Deep Learning and Graph Neural Networks for Mathematical Pattern Recognition. This review highlights the effectiveness of Deep Learning (DL) and Graph Neural Networks (GNNs) in mathematical pattern recognition. Graph-based models, particularly GraphMR built on Graph2Seq, demonstrate superior performance in model accuracy and efficiency over traditional Seq2Seq methods. GNNs effectively handle structured data like ASTs and DAGs, preserving semantic and syntactic information. The integration of encoder–decoder architectures and graph-based reasoning shows significant advancements in recognizing mathematical structures. The evolution from structural methods to DL and GNN approaches underscores the progress in recognition accuracy. As ML adoption grows, the need for large, high-quality datasets becomes critical for training next-generation models.

In graph neural networks (GNNs), pooling operators compute local summaries of input graphs to capture their global properties, and they are fundamental for building deep GNNs that learn hierarchical representations. In this work, we propose the Node Decimation Pooling (NDP), a pooling operator for GNNs that generates coarser graphs while preserving the overall graph topology. During training, the GNN learns new node representations and fits them to a pyramid of coarsened graphs, which is computed offline in a preprocessing stage. NDP consists of three steps. First, a node decimation procedure selects the nodes belonging to one side of the partition identified by a spectral algorithm that approximates the MAXCUT solution. Afterward, the selected nodes are connected with Kron reduction to form the coarsened graph. Finally, since the resulting graph is very dense, we apply a sparsification procedure that prunes the adjacency matrix of the coarsened graph to reduce the computational cost in the GNN. Notably, we show that it is possible to remove many edges without significantly altering the graph structure. Experimental results show that NDP is more efficient compared to state-of-the-art graph pooling operators while reaching, at the same time, competitive performance on a significant variety of graph classification tasks.

Graph neural networks (GNNs) build on the success of deep learning models by extending them for use in graph spaces. Transfer learning has proven extremely successful for traditional deep learning problems, resulting in faster training and improved performance. Despite the increasing interest in GNNs and their use cases, there is little research on their transferability. This research demonstrates that transfer learning is effective with GNNs, and describes how source tasks and the choice of GNN impact the ability to learn generalisable knowledge. We perform experiments using real-world and synthetic data within the contexts of node classification and graph classification. To this end, we also provide a general methodology for transfer learning experimentation and present a novel algorithm for generating synthetic graph classification tasks. We compare the performance of GCN, GraphSAGE and GIN across both synthetic and real-world datasets. Our results demonstrate empirically that GNNs with inductive operations yield statistically significantly improved transfer. Further, we show that similarity in community structure between source and target tasks support statistically significant improvements in transfer over and above the use of only the node attributes.

The Vapnik–Chervonenkis dimension of graph and recursive neural networks