This study aims to investigate factors for the selection of advanced science subject (Ⅱ) in Korean College Scholastic Ability Test (CSAT) employing data mining methods. Using data collected from middle school third year cohort in the Korean Education & Employment Panel (KEEP), we set 1,040 input variables for logistic regression. By applying LASSO and Elastic Net, which are kinds of shrinkage and penalized estimators, we selected and fitted 34 and 86 variables, respectively, in each method. The selected variables were categorized as follows: first, about a half of the variables were from the first year of the survey; second, regional variables; third, variables about friends; fourth, variables describing leisure and hobbies; fifth, traditional variables of students’ individual characteristics; sixth, variables related to school curriculum and career guidance; and seventh, variables related to students’ home environments. It was found, in conclusion, that educational intervention would be effective in resolving the issue of selection of advanced science subject in Korean CSAT.