Unraveling key predictors, mediation pathways, and group heterogeneity of music performance anxiety among Chinese college music students.

Music performance anxiety (MPA) is widespread among music students, underscoring the need to clarify its key predictors, mediation pathways, and group heterogeneity to support its effective intervention. However, existing research often focus on a limited set of theoretically assumed variables for MPA, which restricts the identification of the key predictors among a broader range of potential factors, reducing the accuracy of analyses on group heterogeneity and mediation pathways. To address these limitations, this study analyzed 599 questionnaires from Chinese music students and used machine-learning (ML) models to determine the key predictors of MPA. Based on these predictors, subgroup heterogeneity and mediation pathways were further examined using latent profile analysis and mediation analysis. Incorporating 21 potential predictors, four ML models were developed, with the random forest model demonstrating the highest predictive accuracy for MPA. SHAP analysis further identified emotion-oriented coping (EOC), perceived stress (PS), maladaptive perfectionism (MP), self-efficacy (SE), and social support (SS) as the most dominant predictors for MPA. Based on these five factors, latent profile analysis identified three distinct groups, among which students characterized by higher EOC, PS, and MP alongside lower SE and SS were associated with higher MPA levels. Within the I-PACE framework, mediation analysis further indicated that PS and EOC both individually and sequentially mediated the positive associations between MP and MPA, as well as the negative associations between SS and MPA, representing the primary mediating pathways. These findings highlight the key psychological and social mechanisms underlying MPA, providing evidence-based guidance for interventions aimed at reducing performance anxiety and supporting the mental health of music students.