미용 교육 학습자의 중도포기 의도 예측: 머신러닝 분류 모형의 적용 = Prediction of Dropout Intention among Beauty Education Learners: Application of Machine Learning Classification Models

다국어 초록 (Multilingual Abstract)

As participation in vocational education increases, there is a growing need for in-depth research on dropout prevention in beauty education. This study investigated the association between learners’ perceptions of instructors’ attitudes and dropout intention in beauty education and explored the predictability of dropout intention using machine learning-based classification models. Furthermore, feature importance and SHapley Additive exPlanations (SHAP) analyses were conducted to identify key instructional factors related to dropout prevention. Data were collected through a self-administered survey of learners who were currently enrolled in or had previous experience in cosmetology national technical qualification programs at private beauty academies. The survey was conducted from March 23 to March 31, 2023, and 349 valid responses were analyzed. Logistic Regression and Random Forest classification models were applied in a Python-based analytical environment, and SHAP analysis was performed to interpret model predictions. The results showed that the Random Forest model demonstrated better overall classification performance than the Logistic Regression model, particularly with higher recall and F1-score for Group 1, representing learners with higher levels of dropout intention. Variables representing confidence building, support for independent problem-solving, positive expectations, and instructor trust in learners' chances of passing were repeatedly identified as important predictors. SHAP analysis further confirmed that these variables made meaningful contributions to the prediction of dropout intention. The findings suggest that instructor attitude-related variables, as perceived by learners, may function as important predictors of dropout intention and provide practical implications for the early identification of learners at risk of dropout and the development of educational intervention strategies.