Analysis of factors predicting science high-achievement group

Abstract

This study investigates the predictors of science high-achievement group among Korean high school students, considering raw data of the students’ responses to PISA 2015. In particular, the factors differentiating the excellent group from the general group were analyzed using a random forest analysis with hybrid sampling and hyper-ensemble approach on the responses of 92 individual questions. The results revealed distinct gender-based patterns. Male students captured higher rankings in factors related to science-related reading at the age of 10 and experimental practices. Conversely, female students displayed a stronger correlation between high achievement and non-cognitive characteristics such as connecting the value of science subjects to their future careers. Despite the gender differences in the magnitude of increase in prediction probabilities and variations, their overall direction remained consistent. Notably, both male and female students who were introduced to science-related books around the age of 10 and were presently exposed to scientific literature exhibited a higher likelihood of belonging to the high-achievement group. These findings will aid the development of tailored education support plans for high-achieving students, facilitating the success of these students to foster a scientific potential and brighter future.