New mathematical algorithm developed by Wigner FK researchers can help predict future unexpected events


Researchers at the Department of Computational Sciences at the ELKH Wigner Research Centre for Physics (Wigner FK), Zsigmond Benkő, Tamás Bábel and Zoltán Somogyvári, are developing new computer methods to help find data in the sea of information – i.e. 'big data – that anticipate emergencies and unforeseen situations. Developments in anomaly detection algorithms can help predict unexpected events, such as a heart attack or the next global crisis. A paper summarizing the results of the study has been published in the prestigious international journal Scientific Reports.

Though our daily lives usually unfold in a predictable manner, the unexpected can also frequently occur: we miss our heartbeat during sleep, snow falls in April, or the stock market crashes, the money market collapses. Even if not every unexpected turn can be detected, there are some extraordinary events that also leave traces on the data sets related to our everyday lives. These are often easy to recognize – a stock market crash with price fluctuations that is noticeable to the naked eye, for example. However, at other times the event is much harder to detect, and small differences in an ECG recording, for example, will only be apparent to an expert. Analysts in large data sets use anomaly-detection algorithms to identify such non-obvious changes. Three researchers at Wigner FK have developed such an algorithm that can help predict what the future holds in many areas of life and business].

Hungarian researchers examine unexpected events

Unexpected events are often not only unexpected, but unknown. In other words, they are occurences that we have not experienced before, like a unicorn appearing in front of you in the woods. This means that recognizing them is a serious challenge for automated methods. However, by applying the method of the Wigner FK researchers, which is able to detect new, extraordinary events it will be possible to discover these 'unicorns', meaning unexpected events that appear on a long time series and whose characteristics we do not know in advance.

In our modern world, where information is one of the most important raw materials, this new algorithm can be an important and highly useful tool for recognizing extraordinary and especially previously unknown phenomena. And because the occurrence of an extraordinary event can often predict changes with serious consequences, the new algorithm can be applied effectively in practice in many areas, including finance and healthcare.

Could the 2008 global financial crisis have been predicted with the Hungarian method?

The potential of the method described in this study was first demonstrated by researchers through computer-based tests in which they generated and located anomalies. This enabled them to demonstrate how we can detect unique events that are often invisible to the human eye and that other similar anomaly-detection algorithms fail to uncover. In this case, when the new algorithm was applied to real, measured data sets – where an extraordinary event was already known – it found signs of a short sleep apnea during sleep, for example, on one of the ECG data sets that monitored the heart. This could help doctors take appropriate action in the choice of treatment. Examining historical data on banking transactions, the algorithm also signaled an extraordinary event in the run-up to the 2008 financial collapse in one of the key interest rate indicators, the LIBOR index.

Researchers developing the new method

Dr. Zoltán Somogyvári is a senior researcher at the Wigner Research Centre for Physics, head of the Theoretical Neuroscience and Complex Systems Research Group, and a physicist with a PhD in Neuroscience from Semmelweis University. His main area of ​​research is the development of new data analysis methods to improve our understanding of the functioning of the brain, the causes of certain brain diseases, and the functioning of complex systems in general.

Dr. Tamás Bábel is a physician with a PhD in clinical neuroscience, specializing in medical technology and digital healthcare innovation. His research interests are primarily in neurocardial biomarkers.

Dr. Zsigmond Benkő is a biophysicist with a PhD in theoretical medicine from Semmelweis University. He currently works as a research fellow in the Theoretical Neuroscience and Complex Systems Research Group at the Wigner Institute for Physical Research. Using the tools of artificial intelligence, he develops new data processing algorithms that enable the modeling and control of complex systems.