UWindsor researchers working to predict the unpredictable

Computer science professor Robin GrasComputer science professor Robin Gras can forecast trends in chaotic systems like climate and stock markets.

Two University of Windsor researchers have developed software that can predict the long-term future of complex systems like climate change, financial markets and epileptic seizures.

Abbas GolestantiNature is full of complex systems that are made up of seemingly random events. Until now, that chaos has left us bewildered when trying to predict the future of these systems. But Abbas Golestanti, a PhD student in the School of Computer Science, and his supervisor Robin Gras, an associate professor in the School of Computer Science and Canada Research Chair in Learning and Simulation for Theoretical Biology, have developed a computer model that’s been able to accurately forecast the future of some chaotic systems. Dr. Gras and Golestanti published their findings in Scientific Reports.

“It seems for a given system and a long range of time, the level of chaos stays stable. Even if the value itself is strongly changing, there is some stability in the whole system,” said Gras. “Other existing methods cannot predict trends. Their predictions diverge very fast. This is the first time someone can effectively predict for a long time.”

Gras and Golestani took their model and plugged in two years of data on stock prices in the Dow Jones Industrial Average in an attempt to predict the next six months.

“We show that we would be able to predict a financial crisis even if that crisis does not appear in the actual data we used to train the model,” said Gras.

There is a potentially big medical breakthrough as well. Gras said they are able to predict if someone with epilepsy will have a seizure in the next 17 minutes.

“Up until now the best prediction was a few minutes ahead and had an accuracy of 50 to 60 percent,” he said. “We got 17 minutes with 100 percent accuracy.”

The research team looked at the electroencephalography (EEG) readings of patients. This measures the brain’s electrical activity. A half-hour of readings plugged into the software was enough data to give someone with epilepsy a 17 minutes head’s-up that they were going to have a seizure. Gras and Golestani have patented the software and now they need a device that can constantly monitor a patient’s EEG.

“If you have this device on you all the time, it reads your EEG values and 17 minutes in advance it can ring and say: You will have trouble in 17 minutes. Stop your car. Go to a safe place. Phone someone to help you,” said Gras.

They have also looked at 100 years of climate data and evaluated the next 30 years for climate change. Gras plans to continue the research by looking at the future of earthquakes and heart attacks.