Scholarlink

<p>Fraud detection has garnered increasing attention from the research community due to its critical role in safeguarding online enterprises, such as iQIYI (one of China’s largest online video-sharing and social media platform), against merchant losses resulting from fraudulent activities. Traditional approaches primarily focus on capturing features of users’ historical behavior sequences while neglecting the representation of future behaviors. Consequently, future behavior sequences, rich in informative content, have not been fully leveraged. To address this gap, we propose a novel Fraud Detection framework with Recommendation (FD-Rec), which integrates both historical and future information. Based on our observation that fraudulent users are more likely to replicate actions similar to those exhibited by blacklisted fraudulent users in the near term, we innovatively embed future behaviors through recommendation mechanisms. Furthermore, we introduce a multi-group weighted recommendation strategy that enables user interactions across varying lengths of behavioral sequences. This approach allows for a comprehensive exploitation of co-interactions among users. Extensive experiments conducted on real-world industrial datasets demonstrate the superiority of the proposed FD-Rec framework. After applied the model online, the daily average recall rate has increased by 4% to 5%.</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

Fraud Detection framework integrating recommendation mechanisms for future representation augmentation

  • 期刊论文

  • 发表日期：

    2026年6月2日

查看全文
  • Fangshu Chen,
  • Yixin Tian,
  • Jiahui Wang,
  • Lu Chen,
  • Junqi Pan,
  • Panpan Feng,
  • Huimei Zheng,
  • Surun Ji,
  • Mingfan Lu

摘要

Fraud detection has garnered increasing attention from the research community due to its critical role in safeguarding online enterprises, such as iQIYI (one of China’s largest online video-sharing and social media platform), against merchant losses resulting from fraudulent activities. Traditional approaches primarily focus on capturing features of users’ historical behavior sequences while neglecting the representation of future behaviors. Consequently, future behavior sequences, rich in informative content, have not been fully leveraged. To address this gap, we propose a novel Fraud Detection framework with Recommendation (FD-Rec), which integrates both historical and future information. Based on our observation that fraudulent users are more likely to replicate actions similar to those exhibited by blacklisted fraudulent users in the near term, we innovatively embed future behaviors through recommendation mechanisms. Furthermore, we introduce a multi-group weighted recommendation strategy that enables user interactions across varying lengths of behavioral sequences. This approach allows for a comprehensive exploitation of co-interactions among users. Extensive experiments conducted on real-world industrial datasets demonstrate the superiority of the proposed FD-Rec framework. After applied the model online, the daily average recall rate has increased by 4% to 5%.