‘Can we predict aggression?’—Determining the predictors of aggression among individuals with substance use disorder in China undergoing enforced detoxification through machine learning

•An innovative machine learning method, SHAP, was employed to interpret the important factors that predict aggression.•Psychological security, parental conflict and impulsivity were the top 3 significant predictors of aggression.•There is a significant difference in the predictive effect of depressants and stimulants on aggression.•Alexithymia was one of the top predictors in all models and positively predicted aggression.•Improving emotional regulation and relationships with family and friends may be effective ways to reduce aggression.