Using Machine Learning to Determine a Functional Classifier of Retaliation and Its Association With Aggression

Area of Science:

• Neuroscience
• Developmental Psychology
• Machine Learning

Background:

• Assessing neural system integrity in functional processes is challenging.
• Developing reliable methods to evaluate neural function in adolescents is critical for understanding behavioral development and psychopathology.

Purpose of the Study:

• To evaluate a machine learning classifier for identifying retaliation neural patterns in typically developing adolescents.
• To test the generalizability of this classifier on clinically concerning youth.
• To investigate the association between classifier-determined neural integrity for retaliation and antisocial behavior, proactive, and reactive aggression.

Main Methods:

• Collected blood oxygen level-dependent (BOLD) response data from 82 typically developing and 120 clinically concerning adolescents during a retaliation task.
• Developed a support vector machine (SVM) algorithm using data from typically developing adolescents.
• Tested the SVM classifier's performance on the clinically concerning adolescent sample.

Main Results:

• The SVM classifier achieved high accuracy (92.48%), sensitivity (89.47%), and specificity (93.18%) in distinguishing retaliation phases in typically developing adolescents.
• The classifier demonstrated comparable success in distinguishing neural function in the clinically concerning adolescent group.
• Greater distance from the classifier's hyperplane for retaliation was associated with reduced conduct problems and proactive aggression.

Conclusions:

• This study introduces a preliminary "retaliation endophenotype" using machine learning.
• The functional integrity of this endophenotype is significantly associated with conduct problems and proactive aggression in adolescents.
• These findings highlight the potential of machine learning in understanding neural mechanisms underlying behavioral issues.