Posted: 27 February 2006
Nature teaches computers how to solve real-world problems
UCD’s Natural Computing Research and Applications Group (NCRA) researchers demonstrate how biologically inspired algorithms can help financial modelling
Predicting the future for financial gain is a difficult, sometimes profitable activity. Two leading academics at UCD have turned to nature to develop computer-driven financial models that help develop successful trading strategies and assess credit risk.
According to Dr Anthony Brabazon, "nature has evolved and fine-tuned biological systems over many millions of years. We can draw on this learning by taking ideas from biological systems and combining them with computers to create powerful problem solvers."
Drawing inspiration from a wide array of natural mechanisms including biological evolution, the workings of the central nervous system, the human immune system, and models of social interaction, Dr Michael O’Neill (UCD School of Computer Science & Informatics) and Dr Anthony Brabazon (UCD School of Business) use biologically inspired algorithms for prediction and classification, effectively replicating the step-by-step natural selection process through a computer programme.
These algorithms can be applied in financial modelling and in the development of trading systems. “Financial markets represent a complex, ever-changing, environment in which a population of investors competes for profit. Biological entities have long inhabited such environments, and have competed for resources to ensure their survival. It is natural to turn to algorithms which are inspired by biological processes to tackle the task of survival in a financial jungle.” said Dr O’Neill.
An example of their work is a case using the natural immune system for bond rating classification.
The natural immune system is a highly complex system, comprised of an intricate network of specialised tissues, organs, cells and chemical molecules. The natural immune system can recognise, destroy and remember an almost unlimited number of pathogens (such as viruses, bacteria and parasites). It protects the body by being able to distinguish between self and non-self. Crucially, it does not require exhaustive training with negative (non-self) examples but can identify items as non-self which it has never before encountered. O’Neill and Brabazon draw on the mechanisms of natural immune systems, including their ability to distinguish between self and non-self proteins, to design pattern recognition algorithms that can be used to create a classification system to distinguish between investment and junk rated bonds.
The founders and directors of UCD’s Natural Computing Research and Applications Group (NCRA), Dr Michael O'Neill and Dr Anthony Brabazon, have just published a book on how to apply natural computing to financial modelling. “Biologically Inspired Algorithms for Financial Modelling” is aimed at two audiences: those in the finance community and those in the computer science community and does not require prior knowledge of either biologically inspired algorithms or of financial prediction.
Pictured at official book launch, Friday 24th Feb, 2006: Professor Tom Begley (Head of UCD School of Business), Dr. Michael O'Neill
(UCD School of Computer Science and Informatics), Professor Barry Smyth (Head of UCD
School of Computer Science and Informatics), Dr. Hugh Brady, President, UCD and Dr. Anthony Brabazon (UCD School of Business).
Dr O'Neill points out that "in addition to drawing inspiration from biological evolution in our research, we also examine how populations of people and animals develop social networks, and learn from shared experiences. Social learning plays a vital role in the real world, and we embed this in our computer systems."
Other applications being investigated by researchers at UCD's NCRA include evolving programmes that allow computers to programme themselves, evolving aesthetically pleasing art, and developing systems that can be used for medical diagnosis. The NCRA is undertaking both basic and applied research in natural computing. The developed computer systems can be used in many areas including bioinformatics, telecommunications, operations research, engineering and business.
The website provides more information about the group's activities. UCD’s NCRA research is funded by Science Foundation Ireland and by Enterprise Ireland.