Random Forest Analysis of Factors Predicting Science Achievement Groups: Focusing on Science Activities and Learning in School

Abstract

Abstract This study explored science-related variables that have an impact on the prediction of science achievement groups by applying the educational data mining (EDM) method of the random forest analysis to extract factors associated with students categorized in three different achievement groups (high, moderate, and low) in the Korean data from the 2015 Programme for International Student Assessment (PISA). The 57 variables of science activities and learning in school collected from PISA questionnaires for students and parents were analyzed. Variables related to students’ past science activities, science teaching and learning methods, and environmental awareness were found to played important roles in predicting science achievement. When checking partial dependence plots for major variables, science activities and instructional strategies had a high probability of changing the prediction of an achievement group. This study focused on science-related contextual variables that can be improved through government policies and science teachers’ efforts in the classroom.