Page 47 - CCFA Journal - Second Issue

加中金融

2. 风险管理面临数字化转型 2. Risk management facing digital transformation
近年来，随着互联网经济的快速发展，风险管理数字化相
应开始提速发展。风险管理的数字化是指利用大数据、风 With the rapid development of Internet economy, the
险计量模型和 IT 技术，从超大规模的数据中经济地提取 digitalization of risk management (DRM) has started to speed
风险因子的过程，支持企业的业务决策。风险管理的数字 up. The DRM management refers to the process of extracting
risk factors from super-large-scale data using big data, RM
化，涵盖了所有提高风险效能效率的数字化因素，包含了 model and IT technology to support business decision-making.
流程自动化、决策自动化、数字化监控和预警；风险管理 The D RM covers all the digital factors to improve the
数字化，不仅是传统风险管理技术的信息化过程，更是适 efficiency of risk efficiency, including process and decision
应数字化生态系统发展而进行的风险管理模式创新、流程 automation, monitoring and early warning. It is not only
再造、新技术应用、思维方式改变、运行体系提升的过程。 information process, but also model innovation, process re-
风险管理数字化的本质含义就是在流程、数据、分析、IT creation, new application and way of thinking, and
和组织结构（包括人才和文化）开展协同创新调整。也即 improvement of operating system. The essence of DRM is
指 风 险 管 理 部 门 如 何 应 对 传 统 业 务 的 数 字 化 转 型。 collaborative innovation in processes, data, analytics, IT, and
organizational structures including talent and culture
对于大型金融机构而言，风险管理职能部门必须尽快适应
这一模式的转变，建立起与新模式相匹配的风控措施，支 For large financial institutions, the RM function must adapt to
持业务可持续发展。金融企业需要将数字化浪潮带来的新 the change of above model ASAP. Financial enterprises need to
算法、大数据以及新的技术手段，更好地融入到风控管理 embrace the wave of new algorithms, big data and new
中去，必将为企业创造出新的价值。一方面，金融企业要 technology, and better integrated into the RM which will create
new value for enterprises.
加强数字化技术应用能力，在风险管控中应用大数据与分
析技术，降本增效、创造价值；另一方面，针对数字化带 In the Fintech era, as traditional business models are changed,
来的新风险类型，如模型风险、网络风险等，金融企业要 so will be the RM. The main differences can be seen as to the
提高应对这些新风险的能力。 improvement of business efficiency, such as credit and risk
control efficiency, and to reduce the loss and costs, etc.
在金融科技时代，随着传统的金融业务模式正在被颠覆， Fintech's ability to improve risk control is rooted in multi-
风险管理的方法也将被改变。主要表现为“一增一减”： dimensional, massive and dynamic data. Through the
“增”指提高业务效率，比如信贷效率、风控效率等； integration of own data, open domain data, and third-party data
“减”指减少损失、降低行业成本。金融科技提升风控能 service provider, and channel expansion, one can achieve
力是建立在多维、海量、动态的数据基础之上的。通过对 innovation of leading technologies such as AI, big data and
自有数据整合、公开数据的抓取、第三方服务商合作、以 cloud computing, and apply to the risk model, enterprise credit,
及渠道开发等，实现了人工智能、大数据、云计算等领先 post-credit RM and so on. RM has also changed from the past
技术的创新，并应用到了风险模型、企业征信、贷后风控 manual model to an automated, sophisticated machine
管理等实践领域。风险管理也从过去以人为主的模式，向 identification RM model, in which biometrics, big data analysis,
自动化、精细化的机器识别风险管理模式转变，其中生物 model strategy, machine learning and other technologies are
widely used.
识别、大数据分析，模型策略，机器学习等技术的应用功
不可没。

3. 人工智能等技术推动风险管理数字化 3. New technologies such as AI drive the DRM
目前人工智能和大数据技术的紧密结合已成为风险管理的 The combination of AI and big data technology has become the
核心技术，其基本逻辑是通过在深度学习和数据挖掘中自 core technology to DRM. Its basic logic is to grasp the law of
我更新、自我调整和自我迭代，进而从更多维度的大数据 risks from more dimensions of big data by self-renewal, self-
中把握风险规律。 adjustment and self-iteration in deep learning and data mining.

(1)智能风险扫描 A. Intelligent risk scanning.
In the large-scale risk control scenarios with data complexity,
在数据繁杂的大型风控场景中，运用基于深度学习的人工 the model effect is greatly improved by using AI to generate
智能特征生成框架，对时序、文本、影像等互联网行为、 frameworks based on DL to process and to extract timing, text
非结构化数据深层特征加工提取，大大提升了模型效果。 and image, as well as unstructured data deep features. For
比如消费信贷风险管理通过知识图谱、自然语言处理、机 example, through knowledge map, natural language processing,
器学习等人工智能技术，发现借款人、企业、行业等不同 machine learning and other artificial intelligence technology,
主体间的有效信息维度关联，深度挖掘企业集团、上下游 consumer credit risk management can found more about
合作商、竞争对手、管理人员信息等关键信息。利用大数 borrowers, enterprises, industries and other different subjects
据和人工智能工具，融合外部数据和内部数据，从企业关 for effective information dimension correlation, in-depth
联关系、投资关系、风险要素体系等角度对企业运营状态 mining enterprise groups, upstream and downstream partners,
进行刻画，完成客户关联图谱和信息数据库建设，实时开 competitors, management personnel information and other key
展风险扫描。 information.
Financial institutions need to collect all sorts data of customer,
金融机构需要及时并准确地收集客户数据、交易数据、合 transaction, etc. in a timely and accurate manner, so as to
同数据、市场行情数据、产品数据、财务数据、操作行为 achieve fine classified management. Traditional financial RM
数据、抵质押品数据、机构行为数据等，在此基础上实现 mainly relies on low-frequency data, with the active and
精细化归类管理。传统的金融风险管理主要依赖低频数据， frequent financial transaction activities. Using big data
随着金融交易活动的活跃和频繁，高频数据的获取和处理 technology to track real-time trading data in financial markets
越来越成为微观金融风险管理的重要依据。运用大数据技 such as securities and futures, and to carry out abnormal activity
术跟踪证券、期货等金融市场实时交易数据，并进行异常 analysis, to obtain abnormal orders and entrustment events at
分析，获取具体时点的异常下单、委托事件，能触发交易 specific time points, can trigger real-time early warning and risk
系统或监察系统的实时预警和风险防范。 prevention of trading systems or monitoring systems.

CCFA JOURNAL OF FINANCE DECEMBER 2020
Page 47 第47页