Using XGBoost Techniques to explore predictors of self-management competencies in general high schoo

This study aimed to provide meaningful implications by systematically analyzing variables that predict the self-management capabilities of general high school students using machine learning techniques. This analysis is particularly relevant given the growing emphasis on learner identity in global education. First, to predict the self-management capabilities of these students, eight machine learning regression models—Linear Regression, SVR, KNN, GBM, Bagging Regressor, Decision Tree, Random Forest, and XGBoost—were compared. XGBoost was identified as the most effective model, demonstrating low errors and high explanatory power. Next, the selected XGBoost model was applied to analyze the importance of variables in predicting self-management competency, using the Gain value as the evaluation criterion. It was analyzed that ‘setting a goal (meta-cognition) and studying while thinking about that goal’, ‘(attitude to work) even if it is difficult, I will choose what I like’, and ‘(meta-cognition) I can find and do what I need or task’ are relatively important variables compared to other variables. Additionally, a partial dependence plot was generated to examine the correlation between explanatory variables and response variables. It can be inferred that higher levels of metacognition and a more positive attitude toward work are associated with greater self-management abilities among general high school students.