Dr. Guo and his team designed machine learning models to predict non-suicidal self-injury (NSSI) for Chinese adolescents. 2.5 years of longitudinal data, collected from the Chengdu Positive Child Development Cohort, were broken down into three predictive windows (6 months, 1.5 years, 2.5 years), comparing the performance of several learning algorithms for precision, accuracy, and recall. The team focused on enhancing machine learning interpretability to identify significant risk factors and protective factors of NSSI, and to accurately predict how they interact to influence NSSI developmental trajectories. Guo found that the machine learning model, Random Forests, outperformed consistently across the three windows. SHapley Additive exPlanations (SHAP) analysis showed that suicide-related behavior was the most important predictor across all windows. Moreover, strong protective factors, like spirituality and emotional competence, exhibited clear negative associations with NSSI risk in this study. The researchers claim that the machine learning approach is superior to previous studies in this field because of its methodological advantages: accommodating high-dimensional data and modeling complex interactions across various predictive factors. Still, future studies are needed with larger cohorts in order to strengthen the generalizability in identifying predictors of NSSI.