As the LSEG Tech Talk series continues, the most recent session delved into the realm of applying machine learning and GPUs into the financial markets to develop the ability to detect and adjust for involved risks. The webinar was conducted by two subject matter experts from LSEG, namely Thayaparan Sripavan, the Head of Hardware at Accelerated Systems, and Associate Architect, Janaka Perera.
In order to have a clear idea of the relevance of machine learning in the domain of financial risk management, it is important to spend a bit of time understanding the tendencies of financial markets. In a nutshell, a financial market is a platform where people can buy and sell stocks of companies. However, given the highly dynamic and volatile nature of these markets, there are many risks that need to be assessed and mitigated, with minimal error.
The recently concluded webinar dove into providing an understanding of what financial risk management is, how this job is conducted within financial markets, the current approaches used for assessments and gap identification, and how machine learning algorithms can assist in this process.
How risk is calculated in the world of finance
With that in mind, the webinar began by diving into the kinds of risks that people typically have to reckon with daily. These include market risk, credit risk, liquidity risk as well as operational risk. As an example, credit risk is a scenario where a lender has to bear the risk of whether the borrower will be able to pay back the loan, and this ‘payback ability’ needs to be assessed to reduce overall risk.
In terms of the risk, there are always parameters with crests and troughs that need to be dealt with and it was noted that the LSEG product, Millenium Risk is one of the few services in the market where there is a real-time risk calculation component involved, thanks to the implementation of machine learning processes. This in turn provides customers with greater transparency in terms of their risk position.
Operational risk is also something that needs to be taken into account, given that human error or technical failures can rapidly increase the chances of financial losses. Typically, the approach entails looking at risk models and parameters and what-if analysis is conducted when it comes to risk management.
These processes are generally conducted in a rule-based manner. To use a similar example, if a person was to ask for a loan from a bank, they will be limited by their loan capacity and this capacity will be determined with the aid of a rule. However, in an extremely dynamic environment, it is difficult to make do with simply rule-based approaches as the future state of the system is quite unpredictable.
Enter Machine Learning and GPUs
When traditional models are unable to meet the needs in terms of risk mitigation effectively, risk analysts tend to turn towards machine learning-based solutions aided by GPUs to fill the gaps in the existing systems.
The most significant advantage of machine learning in this domain is their ability to learn from mistakes, making them better problem solvers with time, which cannot be effectively done with the existing approaches. Another massive benefit of machine learning is that there will be a lower cost and lower programmer support needed for the algorithms to generate results.
Machine learning allows financial institutes to mitigate this risk and even avoid the risk altogether. The current laptop will typically, involve GPUs that provide about four to eight processing cores to aid in the creation of models and computing. However, there are GPUs currently available with 10,000 Special Purpose processing cores.
The processing strength of GPUs and their ability to do parallel work bring huge value to machine learning systems. They help bring out the best performance and bring support to the deep learning and machine learning algorithms that are crucial for risk mitigation within financial markets. These GPUs can even be connected together to build major supercomputing processes. During the webinar, Thayaparan and Janaka shared that LSEG has adopted a similar approach when doling out its services.
The promising future ahead
In today’s environment, the implementation of GPUs in the machine learning process has brought several benefits. Ultimately, whether you are looking at assessing the risk of the stock market where billions of dollars move across, or even the risk behind an actual acquisition that may be affected by financial or political changes, this can be benefited by the implementation of machine learning to reduce or mitigate risk. Granted, there’s still some work to be done in terms in terms of regulations and eliminating human bias’ from training datasets. However, the overall results show the adoption of AI in financial markets offers clear benefits for all stakeholders.