The model of machine learning developed by American researchers was able to identify suicidal behavior in adolescents with an accuracy of 91 percent. The article dedicated to the new method is published in Plos One.
Scientists from the University of Brigham Yang collected polls of more than 179 thousand high school students who regularly held in Utah from 2011 to 2017 in order to track drug and mental health consumption. The model created by them and trained on this information was able to predict how the answers a teenager testify to his suicidal behavior, with an accuracy of 91 percent. This is higher than that of existing predictive approaches, the accuracy of which does not exceed 80 percent, emphasize the researchers.
In addition, scientists managed to find questions that have the greatest predictive strength. In particular, among them there were points about Harassment in social networks, bullying schools, quarrels of the house, field, alcohol and marijuana consumption.