Page 15 - CCFA Journal - Third Issue

CCFF2020/加拿大中国金融论坛 2020
Touyz 博士: 首先我很高兴能在这里跟各位专家进行交流。
这是个很好的问题，我想把这个问题的范围先扩大一点，
然后才集中讨论一些与金融机构相关的领域。我喜欢把数
据科学看作是一个流程，通过科学的方法阐明数据的价值。
当我在研究数据科学对金融机构的作用时，它从根本上归
结为我们如何看待通过大数据来驱动价值的工具、方法和
基础设施。这并不一定意味着我们要开始这个极为复杂的
过程，相反，我们要从简单的问题开始，试图把它们弄清
楚，然后在这基础上增加不同层次的复杂问题，直到我们
可以使用更复杂的模型。因此，我们可以利用数据科学先
去解决短期问题，然后去解决更多的长期问题。当我思考
微分值在风险建模和理解用户行为的作用时，是从当前问
题去追溯。数据科学已经被应用到了一些领域，比如第一
资本也在网上发布了许多关于这些领域的内容以及在一些
会议中的新发现。所以数据科学在金融行业中提供的真正
价值是扩展了与金融相关的问题，并在方法上重新定义了
可能。它最终是为客户服务而且通过提供更明智的产品选
择为客户提供差别化的服务。
主持人：那数据如此庞大，你如何使用技术来区分数据的
相关性?
Touyz 博士: 我认为关键是怎样从噪声中分辨出信号。我
想说的是，在更多的行业中，我们现在还不能做到仅仅有
数据（非结构化）就能区分参数、数据集或功能的相关性。
这个需要大量的人工参与和思考，但现在有一个领域正在
尝试用向更自动化的方法去解决这个问题。我对此很兴奋，
而这是一个待解决的问题。
主持人：那接下来让我们将视野从美国银行转到加拿大的
银行。卓女士目前就职于 CIBC 加拿大帝国商业银行，
CIBC 是加拿大最专注于零售的银行之一。那现在多伦多处
于封锁状态，客户怎么去银行、银行怎么联系客户呢？您
的团队是如何帮助银行内部改进数字技术使其能够在新冠
疫情期间联系客户满足他们的财务需求？
卓女士：毫无疑问在新冠病毒期间，客户增加了对数字渠
道的使用。我们也在努力确保客户能够全天候使用我们的
数字渠道。自新冠病毒爆发开始，内部统计数据显示，大
量客户增加了对我们手机银行的使用。数字活动的增加确
实加速了这个行业的转型。我们团队的定位是，在这样一
个困难且具挑战性的时期，为 CIBC 找到持续增长收入的
点，同时保护我们的客户免受欺诈活动的影响。我可以分
享一个我们分析团队是如何努力帮助实现这种向数据渠道
转型的例子。自从新冠爆发以来，我们分析团队开发了一
个机器学习模型以整合数据，包括申请信息，信用卡信息
等，然后工程师和数据科学家根据这些客户信息去增加更
多相关功能。如果在后台发现一个客户开了多个账户，这
将帮助我们确定欺诈的可能性。
主持人：对，安全性是一个很重要的问题，因为现在不能
和客户面对面交流，所有东西都是通过网上进行。
卓女士：这是一个我们内部数据科学家团队利用他们的专
业知识帮助实际应用的很好的例子，但在涉及到更大的系
统集成时，我们可以利用外部技术公司来帮助我们加快和
市场链接的速度。
CCFA JOURNAL OF FINANCE
加中金融
Dr. Touyz: I am so excited to be here, especially among such
esteemed panelists. I'd like to broaden the question a little bit. I
like to think about data science as a process and being able to
articulate the value of the data through the scientific method.
When I think about the differential value in terms of what data
science really can do for things like financial institutions, it
fundamentally comes down to how we think about the tools, the
methods and infrastructure to drive value through things like big
data. That doesn't necessarily mean that we embark on these
incredibly complicated journeys, rather we start with these
simple problems and try to clear them up and then add to these
different layers of complexity until we get to a point where we're
using more sophisticated models. So, there are short term
problems where we can lean in with things like data science, with
the goal of solving some more long-term problems. When I think
about this differentiated value in terms of risk modeling and
understanding user behavior, it comes down to working
backwards from the problem at hand. So, the value that data
science really provides in terms of our industry as a whole is
being able to expand the set of questions that we can ask within
these financial problems, and redefining what is possible in terms
of how our approach. Ultimately, it's in service to the customers
and providing this differentiated value with more informed
offerings.
Moderator: How do you use technology to differentiate the
relevance of data, given that there are so much of them out
there?
Dr. Touyz: I think the key question is how you parse out the
signal from the noise. More broadly within the industry, we're
not at a stage where just because you have (unstructured) data,
you can figure out which one parameter, which set of data or
features are most relevant. That requires a tremendous amount of
human intervention and thought, but there is a field that is trying
to address some of these issues where we're moving to more
automated methods. It's one of the questions that hasn't been
solved yet.
Moderator: Let’s shift from U.S. banks to Canadian banks.
Ms. Zhuo is currently working at CIBC and it is probably one
of the most retail-focused bank in Canada. Currently,
Toronto is in lock down, how do people go to banks and how
do banks reach the customers in a lockdown situation? How
does your team help the bank internally to improve and
enable digital technology to reach out the retail customers
and to meet their financial demands despite COVID-19?
Ms. Zhuo: The pandemic has increased our customers’ use of
digital channels, and we try to make sure the customers can use
our digital channels 24/7. Since the pandemic started, the internal
statistics shows that significantly large number of clients
increased their usage of our mobile banking services, which
really accelerated the transformation in this space. The way we
position ourselves is to identify opportunities for the bank to
continue to grow revenue in such a difficult and challenging time,
but also to help protect our clients from fraudulent activities. I
can share one case of how our analytics team helped enable this
transformation effort. There is an increasing concern in remote
account opening fraud since the start of COVID-19, our analytics
team was engaged during the pandemic period to develop a
machine learning model to consolidate data including
application details, credit card information etc. Then the
engineers and data scientists would add additional features based
on customer information. If we see multiple accounts being
opened this would help us determine the likelihood of fraud.
Moderator: Yes, the security is a key question right now
because you cannot see people face-to-face and everything
is online.
Ms. Zhuo: This is a good example where domain knowledge
from the internal data science team can help us with use cases.
But when it comes to bigger system integrations, we can leverage
external technology companies to help us to speed up the time to
market.
MARCH 2021 Page 15