Paper: arXiv 2502.19305
Authors: Shiqi Wang, Zhibo Zhang, Libing Fang, Cam-Tu Nguyen, Wenzhong Li
Abstract
Corporate fraud detection aims to automatically recognize companies that conduct wrongful activities such as fraudulent financial statements or illegal insider trading. Previous learning-based methods fail to effectively integrate rich interactions in the company network. To close this gap, we collect 18-year financial records in China to form three graph datasets with fraud labels. We analyze the characteristics of the financial graphs, highlighting two pronounced issues: (1) information overload: the dominance of (noisy) non-company nodes over company nodes hinders the message-passing process in Graph Convolution Networks (GCN); and (2) hidden fraud: there exists a large percentage of possible undetected violations in the collected data. The hidden fraud problem will introduce noisy labels in the training dataset and compromise fraud detection results. To handle such challenges, we propose a novel graph-based method, namely, Knowledge-enhanced GCN with Robust Two-stage Learning (\({"\rm KeGCN"}{“R”}\)), which leverages Knowledge Graph Embeddings to mitigate the information overload and effectively learns rich representations. The proposed model adopts a two-stage learning method to enhance robustness against hidden frauds. Extensive experimental results not only confirm the importance of interactions but also show the superiority of \({"\rm KeGCN"}{“R”}\) over a number of strong baselines in terms of fraud detection effectiveness and robustness.
Complexity vs Empirical Score
- Math Complexity: 6.5/10
- Empirical Rigor: 8.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper introduces advanced graph learning techniques (KeGCNR) with mathematical formulations for knowledge graph embeddings and robust two-stage learning, while providing extensive empirical validation on large, real-world financial datasets with public code and detailed experimental results.
Research Flowchart
flowchart TD A["Research Goal: Integrate Rich Interactions<br>for Corporate Fraud Detection"] --> B["Key Inputs: 18-year Financial Graph Datasets<br>(3 Graphs with Fraud Labels)"] B --> C["Challenge Analysis: Financial Graph Characteristics"] C --> D["Challenge 1: Information Overload<br>(Non-company nodes dominate)"] C --> E["Challenge 2: Hidden Fraud<br>(Noisy training labels)"] D --> F["Method: KeGCN_R Model<br>Knowledge Graph Embedding + Robust Two-stage Learning"] E --> F F --> G["Computational Process:<br>1. Knowledge-enhanced GCN<br>2. Stage 1: Representation Learning<br>3. Stage 2: Robust Loss Optimization"] G --> H["Key Findings & Outcomes"] H --> I["1. Interactions are crucial for fraud detection"] H --> J["2. KeGCN_R outperforms strong baselines<br>(Higher Effectiveness & Robustness)"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes Knowledge-enhanced GCN with Robust Two-stage Learning (KeGCN_R), a graph-based model that uses knowledge graph embeddings and a two-stage learning process to handle information overload and hidden fraud.
Source sentence
we propose a novel graph-based method, namely, Knowledge-enhanced GCN with Robust Two-stage Learning (${\rm KeGCN}_{R}$), which leverages Knowledge Graph Embeddings to mitigate the information overload and effectively learns rich representations.
Assumes 3
- Regulators prefer precision in declaring frauds, leading to an asymmetric noise structure where many fraud cases remain undetected (hidden fraud). model
Source sentence
It has an asymmetric noise structure, which is based on the fact that regulators prefer precision in declaring frauds with the cost of letting some fraud cases go undetected (hidden frauds);
- Label noise depends on both the instance and its neighborhood (Instance and Neighbor Dependence). model
Source sentence
It is instance and neighborhood dependence (see Figure 1).
- A company labeled as non‑fraud for at least 8 years can be treated as a clean non‑fraud data point. data
Source sentence
we can reasonably trust a case that has been labeled as non-fraud for at least 8 years and consider it as a clean non-fraud data point.
Data
- Financial graphs from 18‑year records of Chinese A‑share companies fundamentals public
Source sentence
We first extract all the annual financial records of listed companies on two stock exchange markets (Shanghai and Shenzhen) during the period from 2003 to 2020.
The Graph (subgraphs)
Costs
Not applicable
Finds 2
- KeGCN_R achieves the best AUC across all three datasets, outperforming all baselines, especially on the MBM dataset with the most serious hidden fraud problem. performance
Source sentence
our method outperforms all baselines and achieves the best AUC across all the datasets, suggesting that our model can better handle the challenges of corporate fraud detection. Particularly, our method outperforms other methods by a large margin in the MBM dataset, which is the one with the most serious hidden fraud pr
- GNN‑based methods generally perform better than DNNs, indicating that relational information between companies improves detection results. predictive-accuracy
Source sentence
GNN‑based methods generally perform better than DNN, showing that relational information between companies can help improve detection results.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 28 related papers). Coverage is partial, so this is neither replication nor refutation.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (90)
- Heterogeneous graph transformer (2020)
- Multiplex heterogeneous graph convolutional network (2022)
- Financial fraud detection using the related-party transaction knowledge graph (2022)
- Higher-order attribute-enhancing heterogeneous graph neural networks (2021)
- Graph Transformer Networks: Learning meta-path graphs to improve GNNs (2022)
- Graph transformer networks (2019)
- Application of machine learning methods to risk assessment of financial statement fraud: evidence from China (2014)
- Mining corporate annual reports for intelligent detection of financial statement fraud–a comparative study of machine learning methods (2017)
and 82 more
- Financial fraud detection model: based on random forest (2015)
- Deep learning for detecting financial statement fraud (2020)
- Intelligent financial fraud detection: a comprehensive review (2016)
- Detecting financial misstatements with fraud intention using multi-class cost-sensitive learning (2016)
- Detecting fraudulent financial reporting through financial statement analysis (2014)
- Detecting the financial statement fraud: the analysis of the differences between data mining techniques and experts’ judgments (2015)
- The detection of earnings manipulation (1999)
- Predicting material accounting misstatements (2011)
- Board composition and corporate fraud (2004)
- CEO connectedness and corporate fraud (2015)
- The benefits and costs of group affiliation: evidence from east Asia (2006)
- Propping through related party transactions (2010)
- Knowledge graph embedding for link prediction: a comparative analysis (2021)
- Knowledge graph embedding: a survey of approaches and applications (2017)
- Translating embeddings for modeling multi-relational data (2013)
- Complex embeddings for simple link prediction (2016)
- Layer normalization (2016) — link
- A three-way model for collective learning on multi-relational data (2011)
- Knowledge graph convolutional networks for recommender systems (2019)
- Learning entity and relation embeddings for knowledge graph completion (2015)
- Knowledge graph embedding by translating on hyperplanes (2014)
- Convolutional 2d knowledge graph embeddings (2018)
- Holographic embeddings of knowledge graphs (2016)
- DGL-KE: training knowledge graph embeddings at scale (2020)
- Kgat: knowledge graph attention network for recommendation (2019)
- Geometric deep learning: going beyond euclidean data (2017)
- Deeper insights into graph convolutional networks for semi-supervised learning (2018)
- A comprehensive survey on graph neural networks (2020)
- Semi-supervised classification with graph convolutional networks (2016) — link
- Graph Attention Networks (2018)
- Convolutional neural networks on graphs with fast localized spectral filtering (2016)
- Live-Streaming Fraud Detection: A Heterogeneous Graph Neural Network Approach (2021)
- Intention-aware heterogeneous graph attention networks for fraud transactions detection (2021)
- Improving graph-based label propagation algorithm with group partition for fraud detection (2020)
- ANOMALOUS: a joint modeling approach for anomaly detection on attributed networks. (2018)
- Towards consumer loan fraud detection: graph neural networks with role-constrained conditional random field (2021)
- Cash-out user detection based on attributed heterogeneous information network with a hierarchical attention mechanism (2019)
- Heterogeneous mini-graph neural network and its application to fraud invitation detection (2020)
- Heterogeneous graph attention network (2019)
- Heterogeneous graph neural network (2019)
- A survey on heterogeneous graph embedding: methods, techniques, applications and sources (2022)
- Magnn: Metapath aggregated graph neural network for heterogeneous graph embedding (2020)
- Pathsim: meta path-based top-k similarity search in heterogeneous information networks (2011)
- Knowledge embedding based graph convolutional network (2021)
- End-to-end structure-aware convolutional networks for knowledge base completion (2019)
- Modeling relational data with graph convolutional networks (2018)
- Xgboost: a scalable tree boosting system (2016)
- Random forests (2001)
- Least squares support vector machine classifiers (1999)
- Embedding entities and relations for learning and inference in knowledge bases (2015)
- RotatE: knowledge graph embedding by relational rotation in complex space (2018)
- Dimensionality reduction by learning an invariant mapping (2006)
- The case analysis of the scandal of Enron (2010)
- A second-order approach to learning with instance-dependent label noise (2021)
- Learning with instance-dependent label noise: a sample sieve approach (2020)
- Estimating instance-dependent Bayes-label transition matrix using a deep neural network (2022)
- Making deep neural networks robust to label noise: A loss correction approach (2017)
- A Model-Agnostic approach for learning with noisy labels of arbitrary distributions (2022)
- Towards deeper graph neural networks (2020)
- Robust training of graph neural networks via noise governance (2023)
- Nrgnn: Learning a label noise resistant graph neural network on sparsely and noisily labeled graphs. (2021)
- Inductive representation learning on large graphs (2017)
- Fraud detection under multi-sourced extremely noisy annotations (2021)
- On the minimal supervision for training any binary classifier from only unlabeled data (2018)
- Learning with bounded instance and label-dependent label noise (2020)
- Searching to exploit memorization effect in learning with noisy labels (2020)
- Detecting Cash-out users via dense subgraphs (2022)
- Company-as-Tribe: company financial risk assessment on Tribe-Style graph with hierarchical graph neural networks (2022)
- Information transmission between the NASDAQ and Asian second board markets (2004)
- A comparative study of the effects of different factors on firm technological innovation performance in different high-tech industries (2019)
- Learning from massive noisy labeled data for image classification (2015)
- Learning with noisy labels (2013)
- Noise tolerance under risk minimization (2013)
- Classification with noisy labels by importance reweighting (2015)
- Dual t: Reducing estimation error for transition matrix in label-noise learning (2020)
- User behavior pre-training for online fraud detection (2022)
- Part-dependent label noise: Towards instance-dependent label noise (2020)
- Deep learning for credit card fraud detection: A review of algorithms, challenges, and solutions (2024)
- Graph contrastive multi-view learning: A pre-training framework for graph classification (2024)
- Financial fraud detection using graph neural networks: A systematic review (2024)
- Graph Relearn Network: Reducing performance variance and improving prediction accuracy of graph neural networks (2024)
- Cost-Sensitive GNN-Based Imbalanced learning for mobile social network fraud detection (2023)
0 of 90 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.