Scholarlink

<p>The rapid evolution of technology has significantly transformed the financial sector, with credit cards playing a pivotal role in meeting the diverse needs of modern consumers. Credit cards provide a fast and convenient method for making payments and completing financial transactions, enabling consumers to purchase goods and services or access short-term credit. However, credit card fraud remains a substantial challenge, particularly in the context of financial security. Despite the use of artificial intelligence (AI)-based techniques for fraud detection, issues such as concept drift and catastrophic forgetting pose significant barriers to model performance. This paper introduces a Leverage Bagging approach to effectively address both concept drift and catastrophic forgetting in credit card fraud detection. The proposed model leverages the European Union credit card dataset from 2013 and 2023 to uncover changing fraud patterns over time. The model’s efficacy is demonstrated through performance metrics, achieving an accuracy of 99.97%, an AUC-ROC score of 99.97%, and an F1 score of 0.999 on the latest dataset (concept drift), while maintaining 99%, 0.972, and 97% for the older dataset (catastrophic forgetting). These results underscore the potential of the Leverage Bagging approach in enhancing fraud detection in dynamic financial environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adaptive bagging strategies to address concept drift and retain knowledge in fraud detection systems

  • 期刊论文

  • 发表日期：

    2026年2月17日

查看全文
  • Santosh Iranagouda Goudar,
  • K. K. Girish,
  • Biswajit Bhowmik

摘要

The rapid evolution of technology has significantly transformed the financial sector, with credit cards playing a pivotal role in meeting the diverse needs of modern consumers. Credit cards provide a fast and convenient method for making payments and completing financial transactions, enabling consumers to purchase goods and services or access short-term credit. However, credit card fraud remains a substantial challenge, particularly in the context of financial security. Despite the use of artificial intelligence (AI)-based techniques for fraud detection, issues such as concept drift and catastrophic forgetting pose significant barriers to model performance. This paper introduces a Leverage Bagging approach to effectively address both concept drift and catastrophic forgetting in credit card fraud detection. The proposed model leverages the European Union credit card dataset from 2013 and 2023 to uncover changing fraud patterns over time. The model’s efficacy is demonstrated through performance metrics, achieving an accuracy of 99.97%, an AUC-ROC score of 99.97%, and an F1 score of 0.999 on the latest dataset (concept drift), while maintaining 99%, 0.972, and 97% for the older dataset (catastrophic forgetting). These results underscore the potential of the Leverage Bagging approach in enhancing fraud detection in dynamic financial environments.