Obesity remains a persistent global health issue across generations. Targeting family-level factors may help improve child and adolescent body mass index (BMI) outcomes. While associations between parental and offspring BMI are well-documented, the temporal patterns and risk factors driving these relationships remain unclear. This study aimed to identify longitudinal family-level BMI patterns incorporating child, maternal, and paternal BMI and apply interpretable machine learning (ML) methods to uncover key predictors. This longitudinal study included 6092 children and their parents from the TARGet Kids! cohort, with BMI measurements collected from birth to 150 months. Group-based multi-trajectory modeling identified joint trajectories of child BMI-for-age Z-scores (zBMI) and parental BMI. Five ML classifiers predicted group membership using 78 predictors spanning sociodemographic, dietary, parental health, and child lifestyle variables. To explore the modifying effect of parental overweight/obesity (OW/OB) on the relationship between child age and BMI, Bayesian generalized additive mixed models (GAMMs) with smoothed term interactions were applied. Five distinct joint trajectory groups were identified. Children in the highest BMI trajectory group typically had both parents following similar high BMI trajectories. Parental OW/OB status emerged as the strongest predictor of child OW/OB (37 % classification probability), followed by breastfeeding duration (17 %) and child physical activity (15 %). The influence of parental OW/OB was particularly pronounced in early childhood (0-60 months). Bayesian GAMMs confirmed the robust, longitudinal association between child and parental BMI trajectories. Parental BMI patterns strongly influence child BMI development, with age-dependent effects. These findings highlight the importance of early family-based interventions.