Evaluating anomalies and unpredicted events like pandemics and ESG concerns
In part II of the series, we sat down for an interview with Dr. Richard Harmon, Managing Director of Financial Services at Cloudera, to find out more about how the industry is adopting new technology.
You can catch-up and read part 1 of the series, here.
Thank you for joining us for part two of our discussion around data, analytics and machine learning within the Financial Service Sector Dr. Harmon. To start us off, what do you foresee as the biggest opportunities and challenges for Financial Services firms in the next three years?
One of the key takeaways from recent times that should be considered into the future, is that banks need to rethink how they look at tail risk or extreme events that rarely happen.
Machine Learning and AI provide powerful predictive engines that rely on historical data to fit the models. Typically, the more data fed into models, the more robust they become in terms of understanding nuances and subtle relationships. This is critically important for predicting risk exposures. However, another type of analytics, called “prescriptive analytics”, involves simulation tools that look towards the future with a view of many potential scenarios. I am optimistic that what we’ll see over the next 5 years is a mass acceleration and adoption of simulation techniques that complement the continuing innovation in the Machine Learning space.
Prescriptive analytics provides decision-makers with thousands of potential future scenarios. For example, using this information one can evaluate whether something has a set of potential tail risk scenarios that can be catastrophic to the institution or economy, or whether it poses no risk at all. This helps traders, marketing executives, business executives, risk managers, regulators, and others to evaluate what is an acceptable outcome as well as the likelihood of an undesirable outcome. Most importantly, simulation tools do what machine learning algo’s cannot do:
- To take into account feedback effects
- To evaluate scenarios that are outside the limitations of historical data – a key constraint of machine learning models.
- To capture the importance of sequencing of events.
This complementary role of machine learning and prescriptive analytics is the next big trend in AI that we’ll see over the next few years as part of a comprehensive way of helping businesses anticipate change and improve decision making.
How do you see things like Climate Change and Environmental and Social Governance concerns impacting the industry and are there any other major themes you feel could influence how businesses operate in the future?
As you have noted one of the major societal challenges is to address the cause and impact of climate change. More banks and regulators have begun to bring this factor into their future plans – both from an operational perspective as well as their contributions to the local and global communities. An example of this is the growing adoption of Environmental, Social and Governance (ESG) analytics being embedded more and more into investment strategies and in the valuation of equities.
For example, the Government of Singapore is leading by example by helping to standardize company audits taking into account key ESG factors. They see this as helping to bring ESG factors into consideration with the expectation that Singapore institutions will be evaluated favorably given the standards the government has mandated over the past 5 years. This shift to a more ESG-based valuation approach could be a value add or a detriment to company valuations as it will shift investor focus from purely financial metrics to more environmental and social metrics.
Can you talk to us a bit more about simulation techniques, what they mean for the industry, and why they could be important moving forward?
When it comes to prescriptive analytics, one of the newest innovation areas is the use of Agent-Based Model (ABM) simulations. One can think of ABMs as a machine for generating many alternate realizations of the world. This is done via a bottom-up network-based Monte Carlo simulation approach to the modeling of complex and adaptive systems with heterogeneous agents. In the ABM framework, these heterogeneous agents interact with other agents within a network structure (eg., customers of an institution, banks within the financial system, or corporates within an economy).
One important feature of ABMs is that they explain the overall evolution of a system by simulating the behavior of each individual agent – essentially a “digital twin”. This “digital twin” or agent is a self-contained unit that follows a given set of behavioral rules and interacts within many different realizations of possible future environments.
One of the most innovative firms in this space is a fintech company called Simudyne. Their platform runs on top of the Cloudera Hybrid Cloud platform enables one to easily build very sophisticated and massively scalable Agent-Based Models. In financial services, there are many use cases that are optimal for the ABM approach. Some examples include modeling Central Counterparty (CCP) risks, creating adaptive digital markets that allow for realistic testing of algo trading models, creating a richer synthetic dataset to train more robust ML models for various financial crime applications and modeling customer behavior at the individual level with the ability to take into account feedback effects.
Agents in an ABM framework can be driven by very simple rules reflecting their real-world behavioural characteristics or they can be highly sophisticated entities that can learn through the simulations using deep reinforcement learning algorithms.
The complementary ML aspect of ABMs is exemplified by the current Covid-19 environment.
A key consideration with using ML-based models in the current situation is whether this is a ‘structural change’ or a once in a hundred years ‘Tail Risk Event’. If the COVID-19 pandemic is considered a one-off ‘tail risk’ event then when the world recovers, the global economy, the markets, and businesses will operate in a similar environment to the pre-COVID-19 crisis. The challenge in this case is to avoid models from being biased due to the once-in-a-lifetime COVID-19 event.
On the other hand, a ‘structural change’ represents the situation where the pandemic abates and the world settles into a “new normal” environment that is fundamentally different from the pre-COVID-19 world. This “new normal” environment requires institutions to develop entirely new ML models utilizing expanded or alternative data that provides sufficient data to capture this new environment.
The value of the ABM approach is that it provides the ability to support projecting thousands of future scenarios that are not dependent upon historical data limitations – as with ML. This allows decision-makers to evaluate the impact of various shocks, feedback effects, alternative business strategies, and regulatory changes – a key challenge in the current Covid-19 environment.
For more insights into how data and technology are shaping the future of the finance sector, read our latest research on how we’re helping businesses to use data and analytics to fight financial crime and drive business outcome