Automatic detection of fraudulent behavior in networks using graph learning
Autor(a) principal: | |
---|---|
Data de Publicação: | 2021 |
Tipo de documento: | Dissertação |
Idioma: | eng |
Título da fonte: | Repositório Institucional da UFMG |
Texto Completo: | http://hdl.handle.net/1843/41660 |
Resumo: | Graph-based Neural Networks (GNNs) are recent models created for learning representations of nodes (and graphs), which have achieved promising results when detecting patterns that occur in large-scale data relating different entities. Among these patterns, financial fraud stands out for its socioeconomic relevance and for presenting particular challenges, such as the extreme imbalance between the positive (fraud) and negative (legitimate transactions) classes, and the concept drift (i.e., statistical properties of the data change over time). Since GNNs are based on message propagation, the representation of a node is strongly impacted by its neighbors and by the network's hubs, amplifying the imbalance effects. Recent works attempt to adapt undersampling and oversampling strategies for GNNs in order to mitigate this effect without, however, accounting for concept drift. In this work, we conduct experiments to evaluate existing network fraud detection techniques, considering the two previous challenges. For this, we use real datasets, complemented by synthetic data created from a new methodology introduced here. We also propose a new model framework called GMU-GNN, which performs the oversampling of graph nodes belonging to the minority class in order to improve the representativeness and expressiveness in the latent space of features interpreted by the node classification model. In new experiments carried out with 5 datasets, the GMU-GNN obtained a performance superior to the other models currently considered as state-of-the-art under the same contexts and purposes of the problem addressed here. |
id |
UFMG_e4e1ac5df9f4b315cf22ad1d67e88c8d |
---|---|
oai_identifier_str |
oai:repositorio.ufmg.br:1843/41660 |
network_acronym_str |
UFMG |
network_name_str |
Repositório Institucional da UFMG |
repository_id_str |
|
spelling |
Fabrício Murai Ferreirahttp://lattes.cnpq.br/4002187845840872Pedro Olmo Stancioli Vaz de MeloDaniel Sadoc Menaschehttp://lattes.cnpq.br/3346178706128608Ronald Davi Rodrigues Pereira2022-05-13T21:07:22Z2022-05-13T21:07:22Z2021-10-25http://hdl.handle.net/1843/41660Graph-based Neural Networks (GNNs) are recent models created for learning representations of nodes (and graphs), which have achieved promising results when detecting patterns that occur in large-scale data relating different entities. Among these patterns, financial fraud stands out for its socioeconomic relevance and for presenting particular challenges, such as the extreme imbalance between the positive (fraud) and negative (legitimate transactions) classes, and the concept drift (i.e., statistical properties of the data change over time). Since GNNs are based on message propagation, the representation of a node is strongly impacted by its neighbors and by the network's hubs, amplifying the imbalance effects. Recent works attempt to adapt undersampling and oversampling strategies for GNNs in order to mitigate this effect without, however, accounting for concept drift. In this work, we conduct experiments to evaluate existing network fraud detection techniques, considering the two previous challenges. For this, we use real datasets, complemented by synthetic data created from a new methodology introduced here. We also propose a new model framework called GMU-GNN, which performs the oversampling of graph nodes belonging to the minority class in order to improve the representativeness and expressiveness in the latent space of features interpreted by the node classification model. In new experiments carried out with 5 datasets, the GMU-GNN obtained a performance superior to the other models currently considered as state-of-the-art under the same contexts and purposes of the problem addressed here.Redes Neurais baseadas em Grafos (GNNs) são modelos recentes criados para o aprendizado de representações de nós (e de grafos), que alcançaram resultados promissores na detecção de padrões que ocorrem em dados de larga escala que relacionam diferentes entidades. Dentre esses padrões, fraudes financeiras se destacam por sua relevância socioeconômica e por apresentarem desafios particulares, tais como o desbalanceamento extremo entre as classes positivas (fraudes) e negativas (transações legítimas), e o desvio de conceito (i.e., propriedades estatísticas dos dados mudam ao longo do tempo). Como as GNNs são baseadas em propagação de mensagem, a representação de um nó acaba sendo muito impactada pelos seus vizinhos e pelos hubs da rede, amplificando os efeitos do desbalanceamento. Pesquisas recentes tentam adaptar estratégias de subamostragem e sobreamostragem para GNNs a fim de mitigar esse efeito sem, contudo, considerar o desvio de conceito. Neste trabalho, realizamos uma série de experimentos para avaliar técnicas existentes de detecção de fraudes em rede, considerando os dois desafios anteriores. Para isso, utilizamos conjuntos de dados reais, complementados por dados sintéticos criados a partir de uma nova metodologia introduzida aqui. Também propomos um novo framework de modelo denominado GMU-GNN, que realiza a sobre-amostragem dos nós do grafo pertencentes à classe minoritária de forma a melhorar a representatividade e expressividade no espaço latente de características interpretado pelo modelo de classificação de nós. Em novos experimentos realizados com 5 datasets, o GMU-GNN obteve um desempenho superior aos demais modelos tidos atualmente como estado-da-arte sob esses mesmos contextos e propósitos do problema aqui abordado.engUniversidade Federal de Minas GeraisPrograma de Pós-Graduação em Ciência da ComputaçãoUFMGBrasilICX - DEPARTAMENTO DE CIÊNCIA DA COMPUTAÇÃOComputação – TesesRedes neurais (Computação) – TesesDetecção de fraude – TesesFraud DetectionFraudulent BehaviorGraph Neural NetworksAutomatic detection of fraudulent behavior in networks using graph learningDetecção automática de comportamentos fraudulentos em redes utilizando aprendizado em grafosinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesisinfo:eu-repo/semantics/openAccessreponame:Repositório Institucional da UFMGinstname:Universidade Federal de Minas Gerais (UFMG)instacron:UFMGORIGINALM__Sc__Thesis_Ronald.pdfM__Sc__Thesis_Ronald.pdfapplication/pdf1037673https://repositorio.ufmg.br/bitstream/1843/41660/1/M__Sc__Thesis_Ronald.pdfde9291bb7f6a9683ec041cd9aaae579aMD51LICENSElicense.txtlicense.txttext/plain; charset=utf-82118https://repositorio.ufmg.br/bitstream/1843/41660/2/license.txtcda590c95a0b51b4d15f60c9642ca272MD521843/416602022-05-13 18:07:23.101oai:repositorio.ufmg.br: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ório de PublicaçõesPUBhttps://repositorio.ufmg.br/oaiopendoar:2022-05-13T21:07:23Repositório Institucional da UFMG - Universidade Federal de Minas Gerais (UFMG)false |
dc.title.pt_BR.fl_str_mv |
Automatic detection of fraudulent behavior in networks using graph learning |
dc.title.alternative.pt_BR.fl_str_mv |
Detecção automática de comportamentos fraudulentos em redes utilizando aprendizado em grafos |
title |
Automatic detection of fraudulent behavior in networks using graph learning |
spellingShingle |
Automatic detection of fraudulent behavior in networks using graph learning Ronald Davi Rodrigues Pereira Fraud Detection Fraudulent Behavior Graph Neural Networks Computação – Teses Redes neurais (Computação) – Teses Detecção de fraude – Teses |
title_short |
Automatic detection of fraudulent behavior in networks using graph learning |
title_full |
Automatic detection of fraudulent behavior in networks using graph learning |
title_fullStr |
Automatic detection of fraudulent behavior in networks using graph learning |
title_full_unstemmed |
Automatic detection of fraudulent behavior in networks using graph learning |
title_sort |
Automatic detection of fraudulent behavior in networks using graph learning |
author |
Ronald Davi Rodrigues Pereira |
author_facet |
Ronald Davi Rodrigues Pereira |
author_role |
author |
dc.contributor.advisor1.fl_str_mv |
Fabrício Murai Ferreira |
dc.contributor.advisor1Lattes.fl_str_mv |
http://lattes.cnpq.br/4002187845840872 |
dc.contributor.referee1.fl_str_mv |
Pedro Olmo Stancioli Vaz de Melo |
dc.contributor.referee2.fl_str_mv |
Daniel Sadoc Menasche |
dc.contributor.authorLattes.fl_str_mv |
http://lattes.cnpq.br/3346178706128608 |
dc.contributor.author.fl_str_mv |
Ronald Davi Rodrigues Pereira |
contributor_str_mv |
Fabrício Murai Ferreira Pedro Olmo Stancioli Vaz de Melo Daniel Sadoc Menasche |
dc.subject.por.fl_str_mv |
Fraud Detection Fraudulent Behavior Graph Neural Networks |
topic |
Fraud Detection Fraudulent Behavior Graph Neural Networks Computação – Teses Redes neurais (Computação) – Teses Detecção de fraude – Teses |
dc.subject.other.pt_BR.fl_str_mv |
Computação – Teses Redes neurais (Computação) – Teses Detecção de fraude – Teses |
description |
Graph-based Neural Networks (GNNs) are recent models created for learning representations of nodes (and graphs), which have achieved promising results when detecting patterns that occur in large-scale data relating different entities. Among these patterns, financial fraud stands out for its socioeconomic relevance and for presenting particular challenges, such as the extreme imbalance between the positive (fraud) and negative (legitimate transactions) classes, and the concept drift (i.e., statistical properties of the data change over time). Since GNNs are based on message propagation, the representation of a node is strongly impacted by its neighbors and by the network's hubs, amplifying the imbalance effects. Recent works attempt to adapt undersampling and oversampling strategies for GNNs in order to mitigate this effect without, however, accounting for concept drift. In this work, we conduct experiments to evaluate existing network fraud detection techniques, considering the two previous challenges. For this, we use real datasets, complemented by synthetic data created from a new methodology introduced here. We also propose a new model framework called GMU-GNN, which performs the oversampling of graph nodes belonging to the minority class in order to improve the representativeness and expressiveness in the latent space of features interpreted by the node classification model. In new experiments carried out with 5 datasets, the GMU-GNN obtained a performance superior to the other models currently considered as state-of-the-art under the same contexts and purposes of the problem addressed here. |
publishDate |
2021 |
dc.date.issued.fl_str_mv |
2021-10-25 |
dc.date.accessioned.fl_str_mv |
2022-05-13T21:07:22Z |
dc.date.available.fl_str_mv |
2022-05-13T21:07:22Z |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/masterThesis |
format |
masterThesis |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
http://hdl.handle.net/1843/41660 |
url |
http://hdl.handle.net/1843/41660 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
eu_rights_str_mv |
openAccess |
dc.publisher.none.fl_str_mv |
Universidade Federal de Minas Gerais |
dc.publisher.program.fl_str_mv |
Programa de Pós-Graduação em Ciência da Computação |
dc.publisher.initials.fl_str_mv |
UFMG |
dc.publisher.country.fl_str_mv |
Brasil |
dc.publisher.department.fl_str_mv |
ICX - DEPARTAMENTO DE CIÊNCIA DA COMPUTAÇÃO |
publisher.none.fl_str_mv |
Universidade Federal de Minas Gerais |
dc.source.none.fl_str_mv |
reponame:Repositório Institucional da UFMG instname:Universidade Federal de Minas Gerais (UFMG) instacron:UFMG |
instname_str |
Universidade Federal de Minas Gerais (UFMG) |
instacron_str |
UFMG |
institution |
UFMG |
reponame_str |
Repositório Institucional da UFMG |
collection |
Repositório Institucional da UFMG |
bitstream.url.fl_str_mv |
https://repositorio.ufmg.br/bitstream/1843/41660/1/M__Sc__Thesis_Ronald.pdf https://repositorio.ufmg.br/bitstream/1843/41660/2/license.txt |
bitstream.checksum.fl_str_mv |
de9291bb7f6a9683ec041cd9aaae579a cda590c95a0b51b4d15f60c9642ca272 |
bitstream.checksumAlgorithm.fl_str_mv |
MD5 MD5 |
repository.name.fl_str_mv |
Repositório Institucional da UFMG - Universidade Federal de Minas Gerais (UFMG) |
repository.mail.fl_str_mv |
|
_version_ |
1803589346481143808 |