平安壹钱包：大模型赋能风控运营，效率翻倍
InfoQ 讯 随着大模型技术的兴起，金融机构正在积极探索如何利用智能化手段提升风险感知和风控决策能力。平安壹钱包作为国内领先的支付机构，近年来也积极拥抱大模型技术，并取得了显著成果。
据悉，平安壹钱包大数据研发部算法负责人王永合将在即将于 8 月16 日 -17 日举办的 FCon 全球金融科技大会上，分享平安壹钱包如何利用大模型技术实现账户风险管理的数字化转型，以及这一转型如何为金融机构带来实质性的价值。
王永合在接受 InfoQ 采访时表示，作为支付机构，平安壹钱包的业务形态更加多元化，覆盖五大金融增值及消费场景的综合支付服务，包括理财、购物、生活缴费、日常支付以及积分兑换等。因此，平安壹钱包对风控的核心诉求主要集中在全面管控，包括风险的事先预测、实时处置和事后预警监控，以及保护用户交易资金安全。
为了满足这些需求，平安壹钱包构建了风险监控平台和风险运营平台。其中，风险监控平台通过数据分析和机器学习模型对现有案件的风险特征进行重要性分析，生成风控规则，拦截风险交易。而风险运营平台则引入大模型技术，辅助运营人员进行实时诊断和管控。
王永合介绍，平安壹钱包从 0 到 1 建设了智能风控运营平台，该平台主要针对传统风控运营流程中存在的繁琐、专业性强、对抗性高等问题，引入大模型技术进行优化。例如，在“基本信息核查”阶段引入“案件风险点诊断”工具，帮助运营人员更高效地找到风险点；在“电话照会”阶段引入“电话照会剧本生成”工具，根据案件诊断情况生成剧本，让运营人员在电话照会过程中目的更加明确；在“实施管控和撰写小结”阶段引入“管控建议及案件小结生成”工具，评估电话结果，输出诊断建议和管控措施。
通过引入大模型技术，平安壹钱包的智能风控运营平台实现了效率翻倍，有效降低了人工失误率，提升了运营效率，为用户提供了更加安全、便捷的支付体验。未来，平安壹钱包将继续探索大模型技术的应用，不断提升风控能力，为用户提供更加优质的金融服务。
英语如下：
Ping An One Wallet: Large Models Empower Risk Control, Efficiency Doubles!
Keywords: Ping An One Wallet, Large Models, Risk Control
Content:
Ping An One Wallet: Large Models Empower Risk Control Operations, Efficiency Doubles
InfoQ News With the rise of large model technology,financial institutions are actively exploring how to leverage intelligent methods to enhance risk perception and control decision-making capabilities. Ping An One Wallet, a leading domestic payment institution,has also actively embraced large model technology in recent years and achieved significant results.
It is reported that Wang Yonghe, Head of Algorithm at Ping An One Wallet’s Big Data Research and Development Department, will share how Ping An One Walletutilizes large model technology to achieve digital transformation of account risk management and how this transformation brings tangible value to financial institutions at the upcoming FCon Global Fintech Conference, to be held on August 16-17.
In an interview withInfoQ, Wang Yonghe stated that as a payment institution, Ping An One Wallet’s business model is more diversified, covering comprehensive payment services across five financial value-added and consumption scenarios, including wealth management, shopping, utility payments, daily payments, and points redemption. Therefore, Ping An One Wallet’s corerequirements for risk control mainly focus on comprehensive management, including risk prediction, real-time handling, and post-event monitoring, as well as protecting user transaction funds.
To meet these needs, Ping An One Wallet has built a risk monitoring platform and a risk operation platform. The risk monitoring platform analyzes the risk characteristics ofexisting cases through data analysis and machine learning models, generates risk control rules, and intercepts risky transactions. The risk operation platform, on the other hand, introduces large model technology to assist operational personnel in real-time diagnosis and control.
Wang Yonghe introduced that Ping An One Wallet has built an intelligent risk control operation platformfrom scratch, which primarily addresses the issues of tediousness, high professionalism, and strong adversarial nature in traditional risk control operation processes by introducing large model technology for optimization. For example, in the “Basic Information Verification” stage, the “Case Risk Point Diagnosis” tool is introduced to help operational personnel find risk points more efficiently; in the “Telephone Inquiry” stage, the “Telephone Inquiry Script Generation” tool is introduced to generate scripts based on case diagnosis, allowing operational personnel to be more focused during telephone inquiries; in the “Implementation of Control and Writing Summary” stage, the “Control Suggestions and Case Summary Generation” tool is introduced to evaluatetelephone results, output diagnostic suggestions, and control measures.
By introducing large model technology, Ping An One Wallet’s intelligent risk control operation platform has doubled its efficiency, effectively reducing human error rates, improving operational efficiency, and providing users with a safer and more convenient payment experience. In the future, Ping An One Walletwill continue to explore the application of large model technology, continuously enhance its risk control capabilities, and provide users with even better financial services.
【来源】https://mp.weixin.qq.com/s/nLAU9_i5fVxMXdpqLpFo2Q
Views: 1