Computing financial risks

 

COMMERCIAL PROFILE - UNIVERSITY COLLEGE DUBLIN:The group uses advanced computational tools known as evolutionary algorithms – pieces of software which have the capacity to learn from experience

IF A MATHEMATICAL model or computer programme had existed which could tell a bank which customers were good and bad credit risks, we might have avoided many of the problems which have beset the banking sector and our economy over the past two years. Similarly, if a piece of software existed which could enable fund managers to become better stock pickers and asset allocators our pension funds might be in better shape.

If the Science Foundation Ireland funded Financial Mathematics Computation Cluster (FMC2) based at UCD has its way such ideas could become a reality in the not too distant future.

FMC2 is a collaboration between industry, UCD, DCU and NUI Maynooth. It brings together expertise in financial mathematics, financial economics and computer science to create a holistic research programme in financial asset and risk management.

Among the main aims of the cluster is to create a leading centre of research that will underpin the future development of the international financial services sector in Ireland.

“Prior to the mid-1980s Ireland was better known for exports of Guinness than it was for financial services,” explains FMC2 principal investigator Prof Tony Brabazon. “The international financial services sector was virtually non-existent, with most financial services companies focusing on the supply of traditional banking and insurance services to domestic companies and to subsidiaries of overseas multinationals. That changed with the establishment of the Irish Financial Services Centre (IFSC) in 1987. That provided the catalyst for the transformation of the sector.”

Today, more than 450 international institutions operate from Ireland, including more than half of the world’s top 20 insurance companies and more than half of the world’s 50 largest banks. Companies in the IFSC employ some 25,000 people directly and generate a quarter of all Irish service export earnings annually.

Despite this success, we have faced increasing competition from other countries for many of these investments.

A series of discussions was held between the industry and government agencies to identify how Ireland could enhance its attractiveness to the sector. One of the key requirements identified was a transformation of the landscape for education and research in finance in order to generate a flow of suitably trained personnel for companies locating here.

In parallel, the IDA was reviewing financial support offered to international firms wishing to locate in Ireland, redirecting it towards projects focused on RD and product development. This meant that an academic research base was also required.

In order to build this base, a call for proposals was issued by Science Foundation Ireland (SFI) in the area of financial mathematics. This represented the first substantial investment by an Irish science-funding agency in finance-related research: FMC2 was awarded funding of €4.1 million in October 2009.

“Our initial concentration is on the development of theory and methods for the task of asset management,” says Brabazon. “We chose this area as it is one of the primary activities within the IFSC, it offers high value-add and it has substantial growth potential. In addition to its practical significance, activities in asset management encapsulate a wide variety of important scientific research questions in mathematical finance and computation.”

There are two complementary strands to the research. The first concentrates on how portfolios of assets should be constructed and managed. It covers issues such as optimal asset allocation, risk management of the resulting investment portfolios, performance measurement of fund managers, algorithmic trading and efficient execution dealing. The cluster will also develop new software tools for these tasks.

The second stream of research focuses on improving understanding of risk in financial and other investment markets – including property, and the development of better metrics and software tools to manage this risk. It will also examine pension risk in order to develop better methods of managing long-term pension investments.

But the work of FMC2 is a lot more than just “investment by numbers”. The group uses advanced computational tools known as evolutionary algorithms – pieces of software which have the capacity to learn from experience and are the basis of much of the work of the Natural Computing Research Applications (NCRA) group at UCD, a precursor to FMC2.

Natural Computing is the study of computational systems inspired by nature, including physical, social and biological systems. As well developing computational models and problem solving tools, NCRA has a strong tradition of applying these tools to a broad range of problem domains ranging from finance, computer science, design, and architecture to music composition, sound synthesis, bioinformatics and engineering.

Dr Michael O’Neill is a principal investigator with both FMC2 and NCRA. He explains that evolutionary algorithms work in a similar way to the body’s immune system. “The immune system has to recognise patterns,” he notes. “It has to be able to decide whether a cell or molecule is part of your body or not and respond accordingly. It then has to learn which things are threats, and which are not, and so on. If a white blood cell is created which reacts inappropriately it is destroyed – this is how the system learns.”

This hit-or-miss learning process is very important in areas like financial services. “Take credit rating software, for example,” says O’Neill. “This looks at the characteristics of people who have paid you back in the past and suggests lending to others who exhibit similar characteristics. But one of the main traits will be that they have a good credit history already. What happens in the case of a first-time borrower? Other factors must be taken into account to assess whether they represent good risks or not. The software has to learn to tell the difference between different classes of borrowers who might otherwise appear similar.”

It’s also about producing solutions that will survive the test of time, rather than ones that are ‘perfect’. “Perfect solutions are necessarily designed to meet the needs of the environment that exists at the time,” he says. “When conditions change the solution ceases to work. With evolutionary computing lots of solutions are developed that are not necessarily perfect but that will continue to work even as the environment changes – and the solution itself will adapt as well as time moves on.”

The NCRA is already highly successful both in academic and in commercial terms. It has worked with leading companies such as Citi Bank on anti-money laundering measures, and Bell Labs on an innovative self-management system for wireless internet and other networks.

Despite being just one year old and having just completed its second round of recruitment, FMC2 has also built up an impressive commercial track record, having worked with a number of IFSC companies including Pioneer Investments, Ryan Capital and CreditExpo. It is also involved in a collaboration with the Institute of Bankers. The cluster is also in negotiation with three other companies in relation to further projects.

“Manufacturing industry [in Ireland] moved from basic assembly into process innovation and product research and development,” says Brabazon. “The same is happening with financial services. It is moving from basic transaction processing into product innovation. Our aim is to help establish Ireland as a global software development hub for the financial services sector and thus assist in winning new investments for Ireland from this sector.”

FMC2’s other goal is to provide a supply of qualified professionals to the industry in Ireland. “We are bringing together three skillsets here,” Brabazon points out. “Traditional finance, computer science and maths and statistics. People with experience in all three of these areas are required by leading companies. In the academic world the three tend to be siloed away from each other but we are bringing them together to create a research and training environment which can support the industry in the future.”