摘要

In this paper we present the framework for a hybrid model for plastic card fraud detection systems. The proposed data-customised approach combines elements of supervised and unsupervised methodologies aiming to compensate for the individual deficiencies of the methods. We demonstrate the ability of the hybrid model to identify fraudulent activity on the real debit card transaction data. We also explore the model's efficiency against that of the existing monitoring system of the collaborating bank, using appropriate performance assessment criteria.

  • 出版日期2010-8