혼합효과 랜덤 포레스트를 적용한 청소년기 진로성숙도 발달 예측 변인 탐색

Abstract

The purpose of this study is to explore predictive variables of career maturity during middle and high school. To do this, the mixed effects random forest that considers the hierarchical data structure of longitudinal data was applied to GEPS(gyeonggi education panel study) data from the 4th to 9th wave (7th-12th grades). To derive the major predictive variables which are more stable, the analysis through multiple imputation was repeated five times to identify the top 20 variables with a high SHAP (shapley additive expansion) importance index. As a result, 17 major predictive variables were derived by using SHAP value: ‘self-understanding as a learner’, ‘proactive initiative’, sub factors of self-directed learning ability, friend relationships, teacher relationships, and ‘career search during vacation’ were selected as key variables that predict development of career maturity. In addition, variables related to relationships between parent-child such as ‘parent parenting behavior’ and ‘attachment’, as well as students affective attitudes such as ‘self-esteem’, ‘self-efficacy’, and ‘learning motivation’ were selected as key predictive variables, too. The relationship between major predictive variables and career maturity was examined through the SHAP dependence plot, and educational implications for career development of adolescents were discussed.