Result: Lagged predictors of psychological capital in junior high school students: A supervised Machine Learning approach.

Psychological capital (PsyCap)—a higher-order construct comprising hope, self-efficacy, resilience, and optimism—is increasingly studied in educational settings, yet its antecedents remain underexplored. This study aimed to identify longitudinal psychosocial predictors of PsyCap in adolescents using a machine learning framework. We analyzed data from 283 Italian junior high school students (aged ~12–13) who completed validated self-report measures assessing PsyCap and a broad range of psychosocial variables (including positive self-beliefs, dimensions of school motivation, personality traits, individual differences, and school-related social resources) at two time points (T1 = December 2020; T2 = May/June 2021). To predict PsyCap at T2 from T1 variables, we used Elastic Net and Random Forest models, supported by eXplainable Artificial Intelligence (XAI) techniques. Most models achieved <italic>R</italic>² > .60 on the test set, indicating good predictive performance. Beyond the autoregressive effect, the strongest lagged predictors of PsyCap were openness to experience, conscientiousness, and math self-concept. These findings underscore the importance of personality traits and domain-specific positive self-beliefs in shaping PsyCap. Educational implications include the potential for school-based interventions focused on strengthening math self-concept and personality-related learning attitudes to foster adolescents’ psychological resources. Finally, we provide detailed end-to-end Python notebooks for applying this pipeline to other research endeavors. [ABSTRACT FROM AUTHOR]