Educational data mining is used to discover significant phenomena and resolve educational issues occurring in the context of teaching and learning. This study provides a systematic literature review of educational data mining in mathematics and science education. A total of 64 articles were reviewed in terms of the research topics and data mining techniques used. This review revealed that data mining in mathematics and science education has been commonly used to understand students' behavior and thinking process, identify factors affecting student achievements, and provide automated assessment of students' written work. Recently, researchers have tended to use such data mining techniques as text mining to develop learning systems for supporting teachers' instruction and students' learning. We also found that classification, text mining, and clustering are major data mining techniques researchers have used. Studies using data mining were more likely to be conducted in the field of science education than in the field of mathematics education. We discuss the main results of our review in comparison with the previous reviews of educational data mining (EDM) literature and with EDM studies conducted in the context of science and mathematics education. Finally, we provide implications for research and teaching and learning of science and mathematics and suggest potential research directions.