Anomaly detection with the density based spatial clustering of applications with noise (DBSCAN) to detect potentially fraudulent wire transfers

Most anomaly detection models are developed by using expert system methods that mimic human experts. The process to capture the expertise honed by fraud examiners is complicated and practically challenging, often resulting in suboptimal models. This study proposes a clustering-based model that captu...

Descripción completa

Detalles Bibliográficos
Autores: Kim, Yongbum, Vasarhelyi, Miklos A.
Tipo de recurso: artículo
Fecha de publicación:2024
País:España
Institución:Universidad de Huelva (UHU)
Repositorio:Arias Montano. Repositorio Institucional de la Universidad de Huelva
Idioma:inglés
OAI Identifier:oai:ariasmontano.uhu.es:10272/23589
Acceso en línea:https://hdl.handle.net/10272/23589
Access Level:acceso abierto
Palabra clave:Anomaly detection
Fraud detection
Clustering
DBSCN
Density based clustering
Spatial clustering
53 Ciencias Económicas
Descripción
Sumario:Most anomaly detection models are developed by using expert system methods that mimic human experts. The process to capture the expertise honed by fraud examiners is complicated and practically challenging, often resulting in suboptimal models. This study proposes a clustering-based model that captures hidden characteristics of potentially fraudulent wire transfers with less human intervention and expertise. Clustering methods classify and group observations with similar characteristics, excluding anomalies from major clusters. The choice of a clustering method and its parameters is often subjective and significantly affects a set of resulting clusters. In order to reduce the subjectivity of a clustering method while retaining its strength, this study proposes a clustering model with Density Based Spatial Clustering of Applications with Noise (DBSCAN) to detect potentially fraudulent wire transfers of an insurance company. The results show that the DBSCAN models identifies hidden relationships between the variables not only included but also excluded for the modeling with noise wire transfers while less human intervention is needed for clustering parameter selections.