[论文解读] Explainable Deep Behavioral Sequence Clustering for Transaction Fraud Detection
该论文提出了一种名为 FinDeepBehaviorCluster 的新型无监督框架,结合时间注意力双向LSTM进行深度行为序列嵌入,并融合专家生成的特征,以检测欺诈交易。通过利用GPU加速的HDBSCAN变体(pHDBSCAN),该方法在大规模数据上实现了500倍的聚类速度提升,实现了对以往未被发现的欺诈模式的实时检测,返回率为23.52%,且在基于规则的聚类解释中表现出高精度。
In e-commerce industry, user behavior sequence data has been widely used in many business units such as search and merchandising to improve their products. However, it is rarely used in financial services not only due to its 3V characteristics - i.e. Volume, Velocity and Variety - but also due to its unstructured nature. In this paper, we propose a Financial Service scenario Deep learning based Behavior data representation method for Clustering (FinDeepBehaviorCluster) to detect fraudulent transactions. To utilize the behavior sequence data, we treat click stream data as event sequence, use time attention based Bi-LSTM to learn the sequence embedding in an unsupervised fashion, and combine them with intuitive features generated by risk experts to form a hybrid feature representation. We also propose a GPU powered HDBSCAN (pHDBSCAN) algorithm, which is an engineering optimization for the original HDBSCAN algorithm based on FAISS project, so that clustering can be carried out on hundreds of millions of transactions within a few minutes. The computation efficiency of the algorithm has increased 500 times compared with the original implementation, which makes flash fraud pattern detection feasible. Our experimental results show that the proposed FinDeepBehaviorCluster framework is able to catch missed fraudulent transactions with considerable business values. In addition, rule extraction method is applied to extract patterns from risky clusters using intuitive features, so that narrative descriptions can be attached to the risky clusters for case investigation, and unknown risk patterns can be mined for real-time fraud detection. In summary, FinDeepBehaviorCluster as a complementary risk management strategy to the existing real-time fraud detection engine, can further increase our fraud detection and proactive risk defense capabilities.
研究动机与目标
- 解决在金融交易数据中检测稀有、动态且此前未见的欺诈模式的挑战。
- 利用非结构化、高吞吐量的用户点击流序列作为传统风险模型的补充信号。
- 通过结合基于深度学习的序列表征与领域专家直觉,提升可解释性,从而改进欺诈检测。
- 通过算法优化,实现实时、可扩展的数亿笔交易聚类。
- 从高风险聚类中提取人类可读、可操作的规则,用于案件调查与主动风险防御。
提出的方法
- 采用基于时间注意力的双向LSTM,从用户点击流数据中学习无监督的序列嵌入。
- 通过将学习到的序列嵌入与领域专家提供的手工特征相结合,构建混合特征表示。
- 利用FAISS开发了GPU加速的HDBSCAN变体(pHDBSCAN),实现在大规模交易数据上的亚分钟级聚类。
- 通过聚类密度和大小的启发式阈值设定,识别高风险聚类。
- 采用规则提取技术对聚类进行解释,生成条件性、人类可读的聚类模式描述。
- 在真实世界交易数据上,通过分层采样和时间衰减加权,在归纳和直推两种设置下对框架进行评估。
实验结果
研究问题
- RQ1基于深度学习的序列建模能否在无监督条件下有效表征用户行为模式,以实现欺诈检测?
- RQ2结合学习到的嵌入与专家特征的混合特征表示,如何提升欺诈检测性能?
- RQ3GPU优化的HDBSCAN实现能否在数分钟内完成数亿笔交易的可扩展聚类?
- RQ4所提框架在多大程度上能检测到现有实时预测模型所遗漏的欺诈交易?
- RQ5基于规则的聚类解释能否为风险分析师和业务团队生成高精度、可操作的洞察?
主要发现
- 所提出的 FinDeepBehaviorCluster 框架在采样流量中实现了23.52%的返回率,显著优于基线方法。
- 混合特征表示(147维)取得了6.27%的最高F1分数,并节省了11,895.58美元损失,优于纯深度特征与手工特征。
- pHDBSCAN 算法相比原始HDBSCAN实现实现了500倍的速度提升,可在数分钟内完成500万笔交易的聚类。
- 仅使用深度序列特征即实现了70.46%的精确率与1.71%的召回率,展现出对欺诈信号的强大判别能力。
- 案例研究显示,高风险聚类表现出清晰的行为模式,如高速操作、低浏览行为、使用被盗金融工具进行多次高价值购买。
- 从聚类中提取的规则生成了精确的叙事式描述(例如,>10次结账、<1次搜索、使用信用卡、使用qq.com邮箱),与已知欺诈手法(如‘重复作案者’)高度一致。
更好的研究,从现在开始
从阅读论文到最终审阅,大幅缩短您的研究时间。
无需绑定信用卡
本解读由 AI 生成,并经人工编辑审核。