Students from disadvantaged socioeconomic backgrounds generally have worse academic outcomes than their more advantaged peers. However, some resilient students beat the odds and achieve academic success despite socioeconomic adversity. Identifying the factors that promote resilience is of critical theoretical and practical importance. Hence, this study aims to examine the different personal and social-contextual factors that predict resilience. We utilized the 2018 Program for International Student Assessment (PISA) data from Hong Kong and focused specifically on the 1,459 students in the bottom socioeconomic quartile. Of these, 251 were identified as resilient students as they demonstrated a high level of achievement despite being from disadvantaged backgrounds. Machine learning (i.e., random forest classification) was adopted to understand the relative importance of 30 different personal and social-contextual factors in classifying students into those who are deemed resilient versus those who are not. Eight top variables that best predicted resilience were identified, including the use of meta-cognitive strategies, joy of reading, teacher-directed instruction, perception of difficulty of the PISA test, sense of belonging to school, discriminating school climate, self-efficacy, and perceived teacher's interest. This study sheds light on the factors that underpin resilience, providing important theoretical and policy implications.