Heterogeneous Mini-Graph Neural Network and Its Application to Fraud Invitation Detection

Zhu, Yong-Nan; Luo, Xiaotian; Li, Yufeng; Bu, Bin; Zhang, Wenbin; Lu, M.

doi:10.1109/icdm50108.2020.00098

Cited by 11 publications

(19 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Bian et al proposed a Bi-Directional Graph Convolutional Networks (Bi-GCN) to reveal the characteristics of rumors from both top-down and bottom-up propagation [2]. Although the direct adoption of GCN is a tempting solution, previous studies [6], [37], [45] have shown that GCN is inferior in fake news detection tasks due to its incapability of distinguishing different types of relations. The inherent homogeneity of GCN induces challenges for GCN-based models to fully capture the heterogeneity of data associations in social networks.…”

Section: B Graph Neural Network Methodsmentioning

confidence: 99%

“…However, most of the GNNs are designed for homogeneous graphs, in which different types of nodes are treated as the same [13], making them not suitable for modeling the structural representation in the heterogeneous social network. Some methods have been proposed to utilize GNNs to deal with heterogeneous graphs [2], [33], [41], [45]. However, they still have major limitations.…”

Section: Introductionmentioning

confidence: 99%

“…Firstly, many of them are constrained by graph structures and the meta-path structures. Some of them are constrained by a certain type of graph structures (e.g., the type of graphs that can be decomposed into multiple mini-graphs in [45]). Some of them require special/additional definitions of the graph structures, which are difficult for people without sufficient domain knowledge to capture (e.g., the additional connections between special user nodes in [41]).…”

Section: Introductionmentioning

confidence: 99%

“…Secondly, many of these methods do not fully model the heterogeneity in the heterogeneous graph. Nodes are regarded to be homogeneous in some methods (e.g., [45]), while edges are regarded to be homogeneous in others (e.g., [2]) making it difficult to fully model the structural representation in the heterogeneous social network.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Fake News Detection with Heterogeneous Transformer

Li¹,

Sun²,

Hsu³

et al. 2022

Preprint

View full text Add to dashboard Cite

The dissemination of fake news on social networks has drawn public need for effective and efficient fake news detection methods. Generally, fake news on social networks is multi-modal and has various connections with other entities such as users and posts. The heterogeneity in both news content and the relationship with other entities in social networks brings challenges to designing a model that comprehensively captures the local multi-modal semantics of entities in social networks and the global structural representation of the propagation patterns, so as to classify fake news effectively and accurately.In this paper, we propose a novel Transformer-based model: HetTransformer to solve the fake news detection problem on social networks, which utilizes the encoder-decoder structure of Transformer to capture the structural information of news propagation patterns. We first capture the local heterogeneous semantics of news, post, and user entities in social networks. Then, we apply Transformer to capture the global structural representation of the propagation patterns in social networks for fake news detection. Experiments on three real-world datasets demonstrate that our model is able to outperform the state-ofthe-art baselines in fake news detection. 1

show abstract

Section: B Graph Neural Network Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Fake News Detection with Heterogeneous Transformer

Li¹,

Sun²,

Hsu³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…No entanto, os autores de [Pereira and Murai 2021] mostram que essa técnica não tem um desempenho tão bom em datasets mais desbalanceados usando datasets reais e sintéticos gerados pelo simulador de transac ¸ões AMLSim. Outras arquiteturas não tratam do desbalanceamento de classe em tarefas de detecc ¸ão de fraude, mas são projetadas para redes heterogêneas [Zhu et al 2020]. Como diferencial deste trabalho, propomos GNNs treinadas a partir de duas tarefas (previsão de links e do valor transferido) para detecc ¸ão da fraude conhecida como lavagem de dinheiro, utilizando dados reais.…”

Section: Trabalhos Relacionadosunclassified

Delator: Automatic Detection of Money Laundering Evidence on Transaction Graphs via Neural Networks

Assumpção¹,

Souza²,

Lacerda³

et al. 2022

Preprint

View full text Add to dashboard Cite

Money laundering is one of the most relevant criminal activities today, due to its potential to cause massive financial losses to governments, banks, etc. We propose DELATOR, a new CAAT (computer-assisted audit technology) to detect money laundering activities based on neural network models that encode bank transfers as a large-scale temporal graph. In collaboration with a Brazilian bank, we design and apply an evaluation strategy to quantify DELA-TOR's performance on historic data comprising millions of clients. DELATOR outperforms an off-the-shelf solution from Amazon AWS by 18.9% with respect to AUC. We conducted real experiments that led to discovery of 8 new suspicious among 100 analyzed cases, which would have been reported to the authorities under the current criteria.Resumo. A lavagem de dinheiro é hoje uma das atividades criminais mais relevantes, principalmente devido ao seu potencial de provocar grandes perdas financeiras para governos, bancos, etc. Nós propomos o DELATOR, uma nova CAAT (tecnologia de auditoria assistida por computador) para detectar atividades de lavagem de dinheiro baseada em modelos de rede neural que codificam transferências bancárias como um grafo temporal de larga escala. Em colaborac ¸ão com um banco brasileiro, projetamos e aplicamos uma estratégia de avaliac ¸ão para quantificar o desempenho do DELATOR em dados históricos que englobam milhões de clientes. O DELATOR supera uma soluc ¸ão off-theshelf da Amazon AWS em 18.9% em termos de AUC. Conduzimos experimentos reais que levaram à descoberta de 8 novos casos suspeitos dentre 100 analisados, que seriam reportados às autoridades sob os critérios atuais. Introduc ¸ãoLavagem de dinheiro é um termo utilizado para se referir a processos que tentam legitimar dinheiro obtido de formas ilegais, como, por exemplo, tráfico de drogas, sonegac ¸ão de impostos, furtos, etc [FATF 2021]. Tipicamente, a lavagem de dinheiro ocorre através de uma sucessão de transac ¸ões financeiras que ofuscam a origem do dinheiro. Em 2017, o Brasil foi considerado líder mundial em lavagem de dinheiro: a prática foi testemunhada em 23% das companhias nacionais, enquanto a média global era 16%. 1 Entre 2013 e 2017, a Polícia Federal contabilizou mais de R$ 123 bilhões de prejuízos ao país.1 veja.abril.com.br/economia/brasil-e-o-maior-do-mundo-em-lavagem-de -dinheiro/; blog.idwall.co/o-que-e-pld-prevencao-lavagem-dinheiro/

show abstract

The danger of dichloromethane in the synthesis and application of imidazole-based dyes

Liu

Chen

et al. 2022

Dyes and Pigments

View full text Add to dashboard Cite

Heterogeneous Mini-Graph Neural Network and Its Application to Fraud Invitation Detection

Cited by 11 publications

References 19 publications

Fake News Detection with Heterogeneous Transformer

Fake News Detection with Heterogeneous Transformer

Delator: Automatic Detection of Money Laundering Evidence on Transaction Graphs via Neural Networks

The danger of dichloromethane in the synthesis and application of imidazole-based dyes

Contact Info

Product

Resources

About