TRIDENT-Seq: Profile-Matched Prefix-Future Contrast for Label-Efficient Payment Risk Modeling

📄 arXiv: 2610.37365v1 📥 PDF

作者: Weiyu Han, Haoyu Jiang, Jie Sheng, Ruobing Zuo, Huilin He, Ze Zhao, Xiao Tan, Nairong Zhang, Shuai Chen, Tianyi Zhang

分类: cs.LG

发布日期: 2026-10-27

备注: 12 pages, 5 figures


💡 一句话要点

提出TRIDENT-Seq以解决支付风险建模中的标签效率问题

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 支付风险建模 前缀-未来对比 深度学习 模型预训练 金融科技

📋 核心要点

  1. 现有支付风险建模方法在标签稀缺的情况下,无法有效捕捉风险行为的细微差异。
  2. 本文提出的TRIDENT-Seq通过前缀-未来对比方法,利用相似活动轮廓的行为摘要来提升模型的预测能力。
  3. 实验结果显示,预训练的编码器在多个支付业务中提升了平均精度和召回率,并在实际应用中显著降低了损失率。

📝 摘要(中文)

Ant International积累了大量用户和支付业务的交易历史,但确认的欺诈和退款结果稀少且延迟,且特定于每个业务。现有方法无法有效捕捉风险行为的特征。本文提出TRIDENT-Seq,采用前缀-未来对比(PFC)方法,通过匹配每个前缀与相似活动轮廓的后续行为摘要,提升了模型的风险预测能力。实验结果表明,预训练的编码器在多个支付业务中显著提高了平均精度和召回率,且在Ant International的实际部署中,基于金额的退款和整体损失率较之前的XGBoost系统降低了37.7%和37.9%。

🔬 方法详解

问题定义:本文旨在解决支付风险建模中标签稀缺的问题,现有方法往往依赖于大量标注数据,难以有效捕捉风险行为的特征。

核心思路:TRIDENT-Seq通过前缀-未来对比(PFC)方法,匹配每个交易历史的前缀与其后续行为的摘要,从而提升模型对风险的识别能力。该设计旨在通过利用相似活动轮廓的行为信息,增强模型的泛化能力。

技术框架:整体架构包括数据预处理、前缀-未来对比训练和编码器输出的活动预测与分散正则化。模型通过对比学习的方式,优化编码器的表现。

关键创新:最重要的创新点在于前缀-未来对比方法的引入,使得模型能够在标签稀缺的情况下,依然有效地捕捉到风险行为的特征,与传统方法相比,显著提高了模型的预测能力。

关键设计:模型采用了特定的损失函数来优化前缀与后续行为的匹配度,并在网络结构上进行了调整,以适应不同业务的需求。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果表明,预训练的编码器在多个支付业务中提升了平均精度和召回率,具体在Wallet业务中使用10%标签时,超越了全标签的训练效果;在Ant International的实际部署中,基于金额的退款和整体损失率分别降低了37.7%和37.9%。

🎯 应用场景

该研究的潜在应用领域包括金融科技、支付系统和风险管理等。通过提升支付风险建模的效率和准确性,TRIDENT-Seq能够帮助企业更好地识别和预防欺诈行为,从而降低经济损失,提升用户信任度。未来,该方法还可以扩展到其他领域的风险预测和管理。

📄 摘要(原文)

Ant International accumulates large volumes of transaction history across users and payment businesses. Confirmed fraud and chargeback outcomes, by contrast, are scarce, delayed, and specific to each business. A single encoder pretrained on unlabeled histories could serve all businesses, but only if pretraining captures the behavior that separates risk. Coarse activity differences can make contrastive pretraining on transaction logs too easy: transaction volume, amount, and timing can distinguish a history's continuation from arbitrary candidates without modeling how payment resources are used. We present TRIDENT-Seq, a transaction foundation model trained with Prefix-Future Contrast (PFC). PFC matches each prefix to a summary of its subsequent behavior among candidates with similar activity profiles; activity prediction and dispersion regularization act on the encoder output. Across three production payment businesses, the pretrained encoder improves average precision (AP) and recall at a 0.5% intervention budget over architecture-matched scratch training in all six business-budget settings. With a fixed validation set, it exceeds full-label scratch training in mean AP and recall using 10% of the training labels on Wallet and 25% on Shopping and SME. Pretraining only on another business's histories still helps. Replacing profile-matched candidates with random ones cuts recall from 24.62% to 5.64%. In Ant International's deployment evaluation, recorded amount-based chargeback and overall loss rates are 37.7% and 37.9% lower than those of the previous XGBoost-based system under the same intervention budget.