일반고 고등학생의 자기관리역량 유형 분류 및 예측 요인 탐색

Abstract

This study aims to provide basic data to enhance the self-management competency of general high school students and further contribute to the vision of human resource development in the 2022 revised curriculum. To this end, the self-management competency prediction variable was analyzed using RandomForest and its importance was derived. Next, the types of self-management competencies were classified through Latent Profile Analysis. In addition, Multinomial Logistic Regression Analysis was conducted to find out whether the self-management competency predictors derived through Random Forest analysis act as influencing factors for categorized profiles through latent profile analysis. The main results of this study are as follows. First, as a result of analyzing the predictive factors of self-management competency of general high school students by applying the random forest technique, it was found that the variables related to "student competency" had the highest predictive power. A total of seven variables related to "student competency," including "metacognition," "behavior control strategy," and "creativity," were identified as the top 20 variables, and the partial dependence chart analyzed that students with higher student competency have higher self-management competencies. In addition, 'attitude to work' and 'friendship' were also analyzed as major predictors with static linear relationships. In addition, class attitude, academic perception, community consciousness, relationship with teachers, appearance perception, and family perception variables were analyzed as major predictors. Next, as a result of latent profile analysis, the types of self-management competencies of general high school students were classified into four potential profiles as a result of latent profile analysis. Each group showed a consistent pattern of class as a group with a sub-factor of self-management competency above the average, a group with an average level, a group below the average, and a group with a lowest level. Finally, for groups classified through latent profile analysis, the top 20 variables derived through random forest were input in units of 10 to verify their influence through multinomial logistic analysis. As a result of the analysis, it was analyzed that the factors that predict self-management competency have different statistical significance and influence directions according to each profile type classification. The main conclusions based on the results of this study are as follows. First, it was found that it was necessary to apply a differentiated method of enhancing self-management competencies according to the level of self-management competencies of general high school students. In the educational field, customized education considering learners' levels and tendencies, such as achievement level, interest, career, and aptitude, is effective in enhancing students' competencies. Therefore, when diagnosing students' level of self-management competency and presenting optimal variables that comprehensively consider the importance index of random forests and the Odds Ratio of multinomial logistic regression analysis to students at that level, it is judged to be most effective for self-management competency. Second, it was found that the self-management competency of general high school students can be effectively improved by comparison and competency. Variables such as academic-related meta-cognition, behavioral control strategies, and attitude or enthusiasm for classes are also important, but the results of this study show that students' social competencies are comparative and competency areas. Lastly, it was analyzed that the self-management competency of general high school students was greatly influenced by external factors of students. It can be seen that the learning strategy and comparison competency-related variables discussed above correspond to internal variables of students in a large framework. In addition to these variables, among the top variables that statically predict self-management competency, the relationship with the teacher and satisfaction with one's appearance were derived as variables that were statistically significant in the self-management competency profile. These results suggest that in addition to improving self-management competencies caused by internal factors of high school students, self-management competencies can be improved by environmental factors.| 본 연구는 2022 개정 교육과정의 적용을 앞둔 시점에서, 일반고 교육현장의 자기관리역량을 한국교육종단연구(KELS) 7차년도 데이터를 활용하여 체계적으로 진단 및 분석하여 의미있는 시사점을 제공하고자 하였다. 이를 위해 머신러닝 기법 중 상대적으로 안정성과 예측력이 높다고 알려진 랜덤 포레스트를 주요 분석 기법으로 채택하여 예측 요인을 탐색하고, 잠재프로파일 분석을 통해 자기관리역량을 수준별로 유형화하였으며, 다항 로지스틱 분석을 통해 잠재프로파일 유형에 대한 예측 요인의 영향력을 검증하는 과정을 거쳤다. 주요 연구 결과는 다음과 같다. 먼저 랜덤 포레스트 분석 결과 408개의 설명변수 중 ‘(메타인지) 목표를 세우고, 그 목표를 생각하며 공부한다’가 노드 불순도 감소량을 기준으로 자기관리역량에 대해 가장 높은 영향력을 갖는 것으로 나타났다. 그 뒤를 이어 ‘(메타인지) 나에게 필요한 일이나 과제가 무엇인지를 찾아서 할 수 있다’, ‘(일에 대한 태도) 힘든 일이라도 내가 좋아하는 일이라면 선택할 것이다’, ‘(또래애착) 내 친구들은 내 말을 귀 기울여 듣는다’ 등이 자기관리역량에 대해 높은 영향력을 보였다. 다음으로 잠재프로파일 분석 결과 정보지수인 AIC, BIC, SABIC 및 분류의 질을 나타내는 Entropy를 고려하여 잠재프로파일의 수를 4개로 결정하였으며, 각 집단은 자기관리역량 매우 미흡(4.1%), 미흡(22.9%), 보통(51.7%), 우수(21.3%) 집단으로 명명하였다. 마지막으로 다항 로지스틱 분석을 통해 자기관리역량 유형에 대한 예측 요인의 영향력을 검증한 결과 준거집단인 보통집단을 기준으로 우수, 미흡, 매우미흡 집단에 모두 영향을 주는 변수로 ‘(일에 대한 태도) 힘든 일이라도 내가 좋아하는 일이라면 선택할 것이다’, ‘ (또래애착) 내 친구들은 내 말을 귀 기울여 듣는다’, ‘(또래애착) 내 친구들은 나를 잘 이해해 준다’, ‘(공동체 의식) 학급에서 어떤 문제가 발생하면 적극적으로 해결하고자 한다’, ‘(또래애착) 내 친구들은 내 의견을 존중해 준다’, ‘(일에 대한 태도) 어떤 직업을 가지는가가 미래의 나의 인생에 중요한 영향을 미친다’, ‘(개별화) 선생님은 나의 장/단점을 잘 파악하고 계신다’, ‘(메타인지) 내 스스로 학습주제 또는 문제를 찾아내거나 만들어 낸다’, ‘(신체) 나의 용모는 매력적인 편이다’, ‘(창의성) 짧은 시간 안에 여러 가지 새로운 생각을 해낼 수 있다’로 총 10개의 변수가 분석되었다.