East Asian students have consistently performed well in mathematics compared to their international peers. Though many researchers have attempted to explore the factors that underpin their success, most studies have focused only on a limited set of variables. Mathematics achievement, however, is a complex phenomenon, determined by multiple factors, and best understood when different variables across levels of analysis are taken into account. Guided by the socio-ecological theory, the present study attempts to understand the relative importance of individual, microsystem, and mesosystem factors in predicting mathematics achievement. We used the Trends in International Mathematics and Science Study (TIMSS) 2019 data provided by 21,340 eighth-grade East Asian students (Chinese Taipei, Hong Kong, Japan, Korea, and Singapore), 1242 mathematics teachers, and 802 principals. Machine learning (i.e., random forest regression) was used to analyze the data. Results identified 11 key variables that best predicted East Asian students' mathematics achievement. Among these, students' confidence in mathematics, socioeconomic status, and school emphasis on academic success were the top three. Theoretical and practical implications are discussed.