Factors associated with mathematics resilience among lower-ESCS students in Saudi Arabia: PISA 2022 evidence and supervised-learning robustness

PISA 2022 was used to examine factors associated with mathematics achievement and resilience among lower-ESCS students in Saudi Arabia. Primary inference combined student weight, 80 replicate weights, and ten plausible values for 6,928 students in 193 schools. Resilience was defined for each plausible value as membership in the lowest Saudi ESCS quartile with performance at or above the national 75th percentile. Among 1,697 lower-ESCS students in 172 schools, resilient counts ranged from 222 to 256, yielding a pooled prevalence of 14.2% (95% CI 11.8–16.6). Mathematics ease, self-efficacy, ESCS, ICT resources, and disciplinary climate were positively associated with achievement, whereas anxiety, school risk, and bullying were negatively associated. Five-fold school-grouped supervised learning evaluated ten plausible-value-specific targets without sharing schools between training and validation. Random forest performed best for ranking, gradient boosting for threshold balance, and balanced logistic regression for recall. Held-out-school SHAP and permutation analyses prioritized ICT resources, self-efficacy, and anxiety. The geographically distributed sample, summarized across five analytical zones, supports national interpretation and regional planning. The findings also highlight locally testable priorities in confidence, anxiety, safety, classroom climate, and purposeful ICT use.