The influencing factors of non-suicidal self-injury (NSSI) among Chinese adolescents with mood disorders remain insufficiently explored. Leveraging machine learning and interpretable models, this study investigates the key determinants and underlying mechanisms of NSSI in this population. Data were collected in collaboration with multiple hospitals across China. We developed a Random Forest (RF) classifier, achieving a predictive accuracy of 0.78 and an AUC of 0.71, and applied the SHAP method for model interpretation. Results indicate that self-esteem and psychological resilience are the most significant psychological factors contributing to NSSI, while sex, age, and BMI are the most influential demographic variables. Among social and family factors, romantic relationships and overprotective parenting emerged as key predictors. Adolescents engaged in early romantic relationships often experience difficulties in intimate relationships, and romantic setbacks may negatively impact their developmental trajectory. Additionally, adolescents in honor classes exhibit stronger cognitive abilities, which may enhance their adaptability to competitive environments and emotional regulation. These findings provide valuable insights for targeted interventions in clinical and educational settings.