Exploring the Psychological Mechanisms of Self-Aggression in Male Individuals with Substance Use Disorders: A Study Combining Interpretability Automated Machine Learning and Conditional Process Model.

Background: Individuals with substance use disorders often exhibit self-aggressive behaviors, which are linked to childhood trauma, the psychological mechanisms underlying self-aggression in substance use disorders remain unclear. This study utilizes key features extracted based on feature importance from explainable machine learning to establish a conditional process model, exploring significant psychological processes related to self-aggression in males with substance use disorders. Methods: Data were collected from 479 male individuals with substance use disorders in a compulsory detoxification facility. In the first part of the study, all collected variables were analyzed using Auto machine learning. Important features were identified through explainable machine learning analysis. The second part of the study utilized key variables to build a parallel mediation model, based on theories related to self-aggression. Results: In the first part of the study, the model was trained using AutoML, achieving a good prediction result (MAE = 2.338). Through explainable machine learning analysis using SHAP, the top three important features were identified as Meaning in Life, Childhood Trauma, and Sensation Seeking, specifically for males with substance use disorders. In the second part, a parallel mediation model was constructed, incorporating both beneficial and barrier factors. The results showed that Meaning in Life significantly mediated the relationship between childhood trauma and self-aggression, acting as a barrier factor, while Sensation Seeking significantly mediated this relationship as a beneficial factor. Conclusion: The study highlights the mediating roles of Meaning in Life and Sensation Seeking between childhood trauma and self-aggression in males, providing a basis for targeted therapeutic strategies.