El papel de la estadística en la detección del fraude bancario
Loading...
Official URL
Full text at PDC
Publication date
2020
Defense date
07/2020
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
Este documento presenta un primer acercamiento a algunas técnicas estadísticas utilizadas para la detección del fraude bancario.
Aplicando modelos probabilísticos, se pretende etiquetar transacciones como fraudulentas o legítimas. Para este proceso de clasificación a menudo se intenta capturar patrones de fraude. Sin embargo, la constante innovación en las estrategias de engaño dificulta esta tarea. Por este motivo, se introducen también modelos de detección de cualquier comportamiento diferente, para examinarlos con detenimiento después.
Partiendo de una tabla que contiene información sobre miles de transacciones y de modelos de clasificación, ¿se podrá detectar ágilmente el fraude en futuras transacciones?
This document is an approach to some statistical techniques used to credit card fraud detection. Using probabilistic models, transactions are supposed to be correctly classified into fraud transactions and non-fraud ones. This classification problem is often solved by capturing fraud patterns. However, the constant innovation in fraud strategies makes it difficult. Therefore, it is also introduced a model that proposes detecting any anomaly behaviour to check its nature later. Taking thousands of detailed transactions and some classification models, will it be possible to detect credit card fraud quickly?
This document is an approach to some statistical techniques used to credit card fraud detection. Using probabilistic models, transactions are supposed to be correctly classified into fraud transactions and non-fraud ones. This classification problem is often solved by capturing fraud patterns. However, the constant innovation in fraud strategies makes it difficult. Therefore, it is also introduced a model that proposes detecting any anomaly behaviour to check its nature later. Taking thousands of detailed transactions and some classification models, will it be possible to detect credit card fraud quickly?