A quarterly publication of the Autism Research Institute

The Autism Research Review International is quarterly publication of the Autism Research Institute

Fall, 2019 | Number 4, Volume 33

Algorithm can predict aggressive outbursts

Researchers at Northeastern University say they have created an algorithm that can predict the violent outbursts of aggressive individuals with autism spectrum disorders (ASD) up to one minute in advance, allowing caregivers to prevent or prepare for an outburst. 

Matthew Goodwin and colleagues spent 87 hours observing 20 children with ASD and aggression, using a device that helps to predict epileptic seizures (the Empatica E4 wristband) to measure the children’s heart rate, sweat production, skin surface temperature, and arm movements. They then synchronized this data with clocks in biosensors the children wore. This allowed them to correlate each episode of aggression with physiological changes that occurred before, during, and after the episode. 

The researchers say they were able to predict an aggressive outburst one minute in advance with 84 percent accuracy. “As our data set grows and we use more sophisticated machine learning models,” Goodwin says, “I think we might get more than 60 seconds.” Goodwin and his team now plan to test their approach on 240 individuals with ASD and aggressive behavior.


“This wearable device can predict aggressive outbursts in people with autism a minute in advance,” news release, Northeastern University, August 21, 2019.