度小满金融大数据风控模型实践

{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"导读：金融是AI赋能传统行业的重要赛道。本次分享的主要内容为金融大数据风控模型在度小满金融的实践。主要介绍金融大数据风控模型的主要技术方法与在应用层面的主要问题，并结合新冠疫情背景下，探讨下风控模型的发展。"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"金融大数据风控模型的技术方法"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"1. 风险管理中的金融科技"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/e9\/f7\/e9eea1b3e3ca2cf2fae408c1b55606f7.png","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"风险管理中的金融科技主要包括两个方面："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一方面是传统金融的风险管理，包括："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"A卡（Application Scorecard，申请评分卡）"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"B卡（Behavior Scorecard，行为评分卡）"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"C卡（Collection Scorecard，催收评分卡）"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"三张评分卡，是传统金融科技中的一部分。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另一方面是信息技术，包括："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"算法算力的大幅度提升：A（Artificial Intelligence，人工智能）"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"用户的行为数据数字化的存储和挖掘基础：B（Big Data，大数据）"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"强大的资源服务共享能力：C（Cloud，云服务）"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"利用信息技术能力能够有效提升传统风险管理评分卡建模的效果。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"2. 度小满信贷风险"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/da\/2b\/da602ca1587d76b3015474af4f51482b.png","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在度小满信贷业务发展过程中，积累了大量数据和模型相关的实战经验。下面主要介绍关于信贷风险模型在度小满的实战经验。如何去识别信用风险，其核心关键点在于识别借款用户真实合理的资金需求以及评估用户是否拥有较好的还款意愿和能力，主要包括三个方面："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"底层需要的是用户基础画像信息，主要包括用户的年龄、性别、学历、婚姻状况、职业、收入、消费能力、房车等资产以及相应的历史信用信息，金融相比于电商等领域对基础画像的准确度等更为严格，因其涉及了用户真实的还款能力。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"中层为用户的基础行为需求模式，主要是用户当前资金端行为往往与前一段时间内的行为存在较大的相关性，通过这些行为可以预测用户的真实资金需求以及未来的还款表现。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"顶层为用户的社会活动，即“物以类聚，人以群分”，通过关联网络中的聚集性发现欺诈团伙，同时可以发现与还款意愿差关系紧密的人逾期的风险更高，而与还款行为好的用户关系密切的人逾期风险相对较低。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"核心而言，利用用户的基础画像、行为需求，外延到用户社会活动去挖掘用户是否有真实合理的资金需求，是否有较好的还款意愿和还款能力，从而建立区分度良好的风险模型。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"3. 时间序列的处理：贷前"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/69\/f5\/69d5888242a2dc9bb26b0b5fb6c68cf5.png","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"信贷业务通过用户授权获取征信报告，基于征信报告了解用户的信用历史，通过分析用户的行为时序来理解用户的真实现金流需求。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"常见的征信查询包括信用卡审批、贷款审批等，此类查询表征了用户在该时刻的资金需求，通过征信报告中贷款发放情况匹配贷款查询申请时间，可以分析用户资金的用信行为。传统金融行业常利用诸如基于不同时间滑动窗口的加工逻辑方式去进行统计，包括过去一个月三个月、六个月、十二个月，二十四个月征信报告查询次数等、过去一个月、三个月贷款发放笔数等指标。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"征信报告还包括用户的公司地址变更、公积金变更等信息，将上述信息基于时间轴进行划分，就可以对用户在一段时间内的信贷需求和用信情况进行刻画分析。我们利用深度神经网络去进行分析，通过记录时间点、该时间点的动作、该动作的类型以及该动作的多个特征值，将每一项信息转为Item作为输入提供给LSTM单元，获取该单元的输出同时又作为下一个时间序列的输入，不断学习从而获得更好的信息表达。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"相比传统方案，基于这类机器学习的方案能够带来KS 2个点的提升。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"4. 时间序列的处理：贷中"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/9a\/cf\/9a59f8fe5baebc45da9eaf65165795cf.png","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}}]}