Preschool and preadolescent antecedents of externalizing problems: a machine learning approach.

This study used a data-driven approach to elucidate the relative importance of child, maternal, and mother-child dyadic characteristics in predicting externalizing problems across two critical stages, preschool and preadolescence, that mark the development of externalizing problems. Data (N = 1,364) were collected through maternal reports and observations during preschool and preadolescence. Using the random forest algorithm in machine learning, the results showed that the predictive models differed between preschool and preadolescence. For preschool antecedents of externalizing problems, maternal characteristics, such as depressive symptoms, education, and sensitivity, emerged as the most highly ranked predictors, followed by mother-child dyadic characteristics. Moreover, for preadolescent antecedents of externalizing problems, mother-child dyadic characteristics, including conflict and positive relationship, were identified as the top predictors, with maternal characteristics playing a secondary role. While child characteristics were relatively less influential across both age groups, child negative reactivity emerged as a salient predictor during preadolescence. The findings contribute data-driven evidence to elucidate the relative importance of preschool and preadolescent antecedents of externalizing problems, with maternal characteristics playing a central role in early childhood and mother-child dynamics becoming most important during preadolescence. Interventions targeting externalizing problems should be developmentally sensitive, with preschool programs emphasizing maternal well-being and early relational foundations, while preadolescent programs prioritize strengthening the mother-child relationships and addressing dyadic challenges.