最有看点的互联网金融门户

最有看点的互联网金融门户
国际资讯

大数据: 金融领域的变局者

无论从商业还是技术视角来看,“大数据”都引发了很多争议。很难分辨何为真实,何为炒作。大数据为金融机构研发创新性的方案提供了很大的潜力,并可能为其带来显著价值。为了能够攫取价值,金融机构必须借助大数据更好的理解他们的数据,快速获取并且做出有价值的决定。这就要求商业目的、大数据存储和分析方法要一致。

创造价值的一个主要因素是要信任用来做决策的信息。不信任信息源,分析可能就会抛锚;相反,坚定的信任则会帮助其吸引董事会支持大数据分析。随着信息源数量和样式逐渐增多,建立对数据的信任对大数据创造价值变得越来越重要。

金融机构需要建立对大数据的信任并攫取机会。这个报告旨在帮助金融机构理解大数据的技术部分。想要更好的理解金融机构怎么从大数据中最大化利润可以参考:从大数据中攫取价值:在金融领域淘金。

大数据在金融机构中的3V’s

容量、多样、速度——大数据的这3个特征与金融机构息息相关但又充满挑战。他们必须掌握大量数据,但这些数据缺乏结构并且信息源复杂。另外,金融机构需要具有实时处理数据的能力。

容量:大数据就是大

以前,企业用百万兆字节的容量来存储企业数据。现在,存储需要更大的空间,因为公司开始增加存储和调用不同形式的数据:交易和客户明细,贸易数据,博客,音频文件和其他。很多国家的法律规定金融机构必须保存数据10年。历史数据的保存只是对金融机构的要求,这增加了金融机构商业进程的复杂性。一个公司保存的数据可能是音频或者是其他形式的文件,但有一点是必须的,就是10年内这些数据可以被随时调用。

银行和保险公司需要储存、组织和调用大量多样性数据的技术和方法。大数据及其分析可以帮助金融机构利用存储的信息而不仅仅是保存记录。尽管有数据保存的要求,但大多数的金融机构并不会像Google那样必须处理大量的数据。新的大数据技术能够提供解决多样和速度的新方法。

多样:结构缺失和混合源

数据不再是传统的数据形式,也不再只是放在公司的后台部门或者仓库总以备寻找。银行和保险公司正使用多种渠道进行客户互动。客户可以和银行互通邮件,给分支机构打电话咨询,上网搜集信息,用手机进行交易。这就导致数据形式的复杂,不适合传统的表格存储结构。理财顾问必须看大量邮件,了解内容,这样才能找到自己想要的东西。这些结构化和非结构化的数据集合起来就可以了解到客户的全面信息,了解这些不仅需要去后台部门而且得查找客户和金融机构互动的各种方式。

为了能够达到上面说的效果,企业不仅要存储这些信息,还要理解这些信息。这就需要使用搜索技术,这个技术能够搜索非结构性数据,集合这些数据提供有意义的集合给出纳,财富顾问,呼叫中心接线员。最好的信息应该来自于企业内外部的所有构成资产。

多样性的另一方面是要求数据来自组织内部和外部。以前,银行从其集团内部获取所需信息。但是,随着外部相关信息的增加,银行和保险公司需要整理内外部信息,并及时处理这些信息。这就需要强大的内容管理工具。

例如,在法国,公众可以通过www.data.gouv.fr接触到近300000个数据点,这些数据包括国家、公众和政府的各方面数据。越来越多的公司能够使用这个外部信息来营销和服务客户。想想YAGO这种信息源(在萨尔布吕肯的马克思普朗克计算机科学研究所的知识库)。2012年,YAGO2s有超过1000万的公司条目,包括关于这些实体的1.2亿的事件。YAGO的这些信息是从Wikipedia、Wordnet和Geonames自动提取出来的。在一个样本中,YAGO的准确率被评估为95%以上。整个数据库的克隆件、专题及其子集都是可用的。它还可以通过不同浏览器查询。YAGO已经被用于IBM的沃顿人工智能系统。

另外,社会媒体不仅为商家提供了外部信息,同时也有助于督促公司更好的服务客户和改善产品。金融机构可以利用客户在企业外部交流时对产品的评价。例如,银行可以从客户的外部博客和网上客户论坛寻找信息。越来越多的银行和保险公司不得不用云方法把他们的网络和客户社区相连。通过倾听客户和收集反馈,他们可以开发出更快、更贴近客户需求的产品。

UFRQTT(NGU@3FD8G66JR~VX 第1张

速度:来自真实世界的实时数据

数据不再只是关乎编译和处理了。它是来自真实世界的实时数据。美国地震调查局地震推特调度(TED)通过用多种语言分析推特的实时容量,能够用时间、地理标记和推特中的位置数据在30秒到2分钟内精确指出地震点。TED提供的地震警报时间比地震仪确认的信息提前了一半。金融机构现在可以使用实时分析手机和社会数据等分辨出欺诈和正常的信用卡消费。

当世界进入电子交换时代,银行和保险公司需要使用多种渠道进行客户互动。每种通道都需要实时进入客户信息以支持互动,每个接触点都对客户有了新的了解。

每年,很多芯片都被用来收集数据以抵抗欺诈和其他状况。大数据已经改变了保险定价,保险公司可以把芯片放入汽车来收集汽车的信息。保险公司可以根据汽车行为和客户驾驶模式来定价。另一个机遇是用交通、天气和地理数据等信息结合汽车和司机来了解情况。

银行和保险公司需要使用实时数据做商业决定的新能力。例如,银行以前要花3到4天来处理信用申请。现在,一些银行就可以在24小时内完成风险评估程序并反馈给客户。保险公司在5分钟内就可以提供报价及其与其他保险公司的价格对比。你可以通过手机进行保险索赔,包括事故车辆损坏的照片、路边援助、GPS坐标等等。将来,通过提高实时数据分析的能力,银行将不再担心巴塞尔流动性风险,因为它能够在每次信贷决定时基于数据实时重算。

将来会管理哪种数据?可能会提供哪种服务?

传统的数据模型基于样本,因为模型工具不能立即处理所有数据。因此,由于样本偏差带来的潜在错误受到普遍担忧。然而,大数据技术能够处理模型中的所有数据,处理成百上千个情景组合,避免了样本偏差。另外,大数据的技术工具不仅处理大容量,而且处理即时数据,也就是说,可以在很短的时间内进入并处理大量数据。基于规则引擎技术使公司可以设计自己的商业处理规则。这些规则允许处理器自助决定85%到90%的事件。

2 第2张

价值计算

解决容量、多样和速度的技术工具使金融机构建立组织信任和处理大数据创新成为可能。为了能够从大数据中攫取最大价值,金融机构应该通过他们要完成的商业目的开始他们的大数据活动。商业目的应该统治大数据应用的领域。这将有利于确定数据收集的范围,防止收集无用的数据。

现有和未来的法规也是金融企业大数据活动的关键因素。这些法规更多的重点放在公司增加监管、透明度和风险报告上面,这远远超越了以前的要求。

除了遵守法规,金融机构应该将商业目的作为数据管理的严格标准。这样的标准将会帮助公司把大数据转化为价值。

从技术角度,大数据可以根本上降低技术方案持有者的总成本。大部分的大数据技术依赖便宜的商业硬件,所以扩大规模很迅速而且经济。他们也会使用开源软件以避免许可费。这方面降低了整个组织采用大数据分析的壁垒。

更广泛点说,大数据的真实价值很难估计。当部分价值可知的情况下,可以通过数学假设和计算来定义价值。如果结构正确,可以用ROI来计算大数据价值。它也可以用来决定新软、硬件购买平衡点、资源配置和项目优先级。

为了减少资本支出和风险,很多金融机构选择了云技术。端到端的云技术分析(例如凯捷的Elastic分析)能让金融机构利用消费模式的所有优势,感觉就像是在经营自己的数据中心。对于很多正在和数据泛滥斗争和需要对用户新需求做出迅速反应的金融机构,云技术可以让其更有效的进入端对端商业智能和大数据分析方案。

银行、资本公司和保险公司都意识到企业信息是最有战略价值的资产。一些金融机构已经把数据库转移到了企业层面的程序,把信息提升到了战略资产的地位。他们正在把公司重塑为信息中心的企业,领导团队管理着消费数据及其质量。但是大部分的公司也说他们没有重点、技能、能力和领导力来有效的管理这个战略资产。对于很多企业来说,最大的大数据问题就是“我怎么使用它呢?”

From both a business and technology perspective, “big data” is generating a lot of discussion—so much so, that it can be hard to know what is hype and what is reality. Big data holds tremendous potential for financial services firms to develop new and innovative solutions that result in significant business value. To capture that value companies must leverage big data solutions to make better sense of the real data they have, get to it quickly and make valuable decisions. It requires congruence between business objectives and the big data storage and analytics approach.

A major factor in the creation of value is trust in the information used to make decisions. Lack of trust in the information sources and analytics can derail the success of an analytics project and, conversely, solid trust can be a huge benefit in appealing to executives and boards of directors for funding of big data analytics initiatives. Establishing trust in big data is paramount to value creation as the variety and number of information sources grows.

What do financial services (FS) companies need to know to establish trust in big data and drive the right business opportunities from it? This paper focuses on helping FS companies to understand the technology component of big data value generation.  To better understand how companies can maximize the value they generate from big data from the business perspective, see Grabbing Value from Big Data: Mining for Diamonds in Financial Services.

The 3V’s of Big Data in Financial Services

The three common characteristics of big data—volume, variety and velocity—are both relevant and challenging for FS companies. They must maintain a lot of data  over time, the data lacks structure, and it is of mixed origin. In addition FS companies  need the ability to process real data in near real-time.

Volume: Big Data comes in one size: Large!

For years enterprises planned terabytes of storage for enterprise data. Now, more and more frequently, storage needs are described in petabytes and exabytes as companies increasingly store and retrieve data of all forms: transactions and customer details, trading data, telemetrics, weblogs, audio files and more.By law in many countries financial services firms must be able to recover data for 10 years. This requirement for historical information is unique for the financial services industry and adds complexity to FS firms’ business processes. It could be an audio file or any form of documentation and a firm must be able to retrieve it for a decade.

Banks and insurers need technologies and methods to store,organize and retrieve a new volume and variety of data. Yet big data and analytics offer new business opportunities that leverage stored information far beyond record retention. Despite retention requirements most financial services firms are not working with the petabytes of data a company such as Google must handle. New big data technologies are able to deliver innovative solutions that deal with variety and velocity,even when very large volumes are not yet present.

Variety: Lack of structure and mixed origin

Data is no longer defined by traditional data types or found in traditional data warehouses or back offices. Banks and insurers are using multiple channels for their customer interactions. A customer can exchange email with the bank, call the branch with  questions, gather information online and conduct transactions on a mobile phone. This results in a multitude of data types that don’t fit traditional tabular (row,column) data structures. Wealth advisors must have access to large volumes of emails, know what’s in these emails, and be able to search them to find the data they need. The structured and unstructured data when integrated offer a 360-degree view of customers—and enable access to that comprehensive information not just in a database in the back office but in all interactions the customer is having with the FS company’s channels.

To support the above scenario the enterprise must store these communications but also understand the content of each one. This requires use of search engine technology (such as Natural Language Processing and Text analytics) that gives the company the ability to search unstructured data,aggregate this information and provide meaningful integration to present to the teller, wealth advisor or call center operator the best information updated with all assets inside and outside the enterprise.

An additional dimension of variety is that data comes from both inside and outside the organization. Historically a bank has drawn its reports and information needs from data that resides inside the organization. Yet, with ever increasing volumes of relevant information now residing outside the company, banks and insurers are challenged to manage inside and outside sources and marshal that information in a timely and relevant manner. It requires powerful content management tools to do this.

For example, in France 300,000 data sets are now available to the public on www.data.gouv.fr, including statistics about all aspects of the country, its people and its government. More and more companies are able to use this outside information to help market their business and service their customers. Consider sources such as YAGO1, a knowledge base developed at the Max Planck Institute for Computer Science in Saarbrücken. As of 2012, YAGO2s has knowledge of more than 10 million entities such as corporations and businesses and contains more than 120 million facts about these entities. The information in YAGO is  extracted automatically from several sources like Wikipedia,Wordnet and Geonames. The accuracy of YAGO was manually evaluated to be above 95 percent on a sample of facts. Copies of the whole database are available, as well as thematic and specialized subsets. It can also be queried through various online browsers. YAGO has been used in the Watson artificial intelligence system of IBM.

In addition, social media is giving companies data from outside the company for their own business intelligence but also represents a source of information for the company to better service customers and improve product innovation. Financial services firms can leverage what the consumer is saying about their product outside the company with other customers. For example, a bank may direct a customer to an outside blog or online customer community for more information. More and more banks and insurers will have to manage their networks and communications with communities of customers using cloud solutions that connect to social media. By listening to these customers and gathering peer-to-peer product feedback they will be able to innovate products more rapidly and better meet customer needs.

UFRQTT(NGU@3FD8G66JR~VX 第1张

Velocity: Coming from the real world in real-time

No longer is data something that is compiled and processed.It is real-time from the real world. By analyzing volumes of real-time tweets in multiple languages, the United States Geological Survey’s Twitter Earthquake Dispatch (TED) is able to use time,  geo-tagging and location data contained within the tweets to pinpoint earthquakes in anywhere from 30 seconds to two minutes. About half the time, TED provides earthquake alerts before seismometers can confirm them. Financial services firms can now  conduct real-time analytics on a variety of sources such as mobile and social data to distinguish between fraudulent and normal credit card activity.

As the world moves toward digital transformation banks and insurers need to use multiple distinct channels for customer interaction to grow their business. Each channel needs real-time access to customer information to support the interaction and each touch point adds new information to the customer’s profile.

Every year more chips are being put in products letting companies collect data to combat fraud as well as a multitude of other possibilities. Big data has changed insurance pricing since Insurance companies began putting a chip in cars to collect information on what the car is doing. Insurers can now offer incentives based on the behavior of the car and customer driving patterns. The next opportunity is to overlay other data sources such as traffic, weather and geographical data to gain greater insight into a specific car and driver combination.

Banks and insurance companies need new capabilities to make business decisions with real data in near real time. For example, banks used to take three to four days to respond to credit applications. Now some are completing risk management processes and getting responses to customers in 24 hours. Insurers are providing quotes in five minutes along with a price comparison against other firms. More and more you can initiate an insurance claim on a mobile device,including taking a photo of the damage from a car accident, getting roadside assistance, transferring the GPS coordinates for the claim and more. In the future, it’s likely that by harnessing real-time analytics capabilities banks will not have to worry about Basel liquidity risk because it could be recalculated in real-time with each credit decision based on data that is in memory and accessible in nanoseconds.

What kind of data will you manage tomorrow? What kind of services will you be able to deliver?

Traditional data modeling is typically based on a data sample because the modeling tools used cannot handle all the data at once. As a consequence, potential errors due to sampling biases, is a common concern. However, big data technologies can handle entire data sets in models and churn through hundreds of scenario combinations and thus help avoid a sampling bias. Furthermore, technology tools for big data are not only about processing big volumes. They also process data that is in memory, which means it can be accessed in nanoseconds and process large volumes in a very short time. Rules engine-based technologies enable companies to design into processes business rules that will allow the process to decide itself (without human intervention) the decisions for 85 to 90 percent of cases.

2 第2张

Calculating Value

Volume, variety and velocity define what the technology toolset must address to enable FS businesses to establish organizational trust and harness the innovation possibilities of big data. To maximize value from big data, FS companies should begin their big data initiatives by agreeing on the business objectives they are trying to accomplish.  Business objectives should govern the area where Big Data should apply. This helps to scope the data collection effort and prevents gathering and managing data that is not needed or usable.

Existing and upcoming regulations are also a key driver for much of the big data activity within FS companies. These regulations are putting greater emphasis on firms to increase governance, transparency and risk reporting, driving the need to go beyond traditional data analysis.

Beyond what is required for regulatory compliance, FS companies should define their own customer-focused data policy to serve as a strict guideline for data management.  Having such a policy in place will help to achieve big data’s intended value.

From a  technology perspective, big data has the potential to substantially lower the total cost of ownership of technology solutions. Most big data technologies rely on inexpensive, commoditized hardware and therefore scale rapidly and very economically. They also make use of open source software avoiding licensing fees. This aspect in turn lowers the barriers to adoption and incorporation of analytics throughout an organization.

More broadly, finding the actual financial value of the data (ROI) can be a challenge. Value definition can be done using mathematical assumptions or real calculations where the component values are available. If  structured properly, the ROI computation can be used to determine the big data value in respect to the data owner’s relative  configuration and monetary data asset value, thus removing the arbitrary aspect of “It Depends.” It is also beneficial to determine the breakeven price point on the purchase of new hardware and software, resource allocations, and project priorities.

To reduce capital expenditures and risk, many FS firms are looking to the Cloud. End-to-end Cloud-based analytics solutions such as Capgemini’s Elastic Analytics enable FS firms to take full advantage of a consumption-based model while maintaining the same look and feel to applications as if they were running in their own data centers. For organizations that are struggling with the deluge of data and the ability to rapidly respond to new demand for insight from their business users, Cloud offers FS firms a way to access an end-to-end Business Intelligence and Big Data Analytics solution with a much shorter ‘time-to-value’.

Banks, capital market firms and insurance companies all realize today that enterprise information is one of their most strategic corporate assets. Some financial institutions have already embarked on a path to transform their disparate operations and data repositories into an enterprise-level program that elevates information to its deserved status of strategic asset. They are reshaping the business into a truly information-centric enterprise where both data quality and consumption are aggressively and consistently managed by the leadership team. But most organizations also agree that they lack the focus, skills, competency and leadership to manage this strategic asset as effectively and efficiently as they would like. For many, the biggest big data problem they face is “How do I use it?”


用微信扫描可以分享至好友和朋友圈

扫描二维码或搜索微信号“iweiyangx”
关注未央网官方微信公众号,获取互联网金融领域前沿资讯。

发表评论

发表评论

您的评论提交后会进行审核,审核通过的留言会展示在下方留言区域,请耐心等待。

评论

您的个人信息不会被公开,请放心填写! 标记为的是必填项

取消

王 少龙

2
总文章数

TA还没写个人介绍。。。

互联网金融狂潮简史

秋源俊二 6小时前

[未央研究]本周互联网金融回顾 | 2017年第46周

未央研究 22小时前

他的模式全球风靡,却在中国水土不服?

一本财经 1天前

蝶变与重生:大变革下的金融科技创新风口

孟永辉 2天前

未央今日播报:天津下发互金消费者权益保护指导意见 东盟六国建立金融科技产业联盟

未央研究 2天前

版权所有 © 清华大学五道口金融学院互联网金融实验室 | 京ICP备17044750号-1