摘要：Objectives: The purpose of this study is to evaluate the factors of first graders’ school adjustment in Korean elementary school by using latent profile analysis and machine learning to predict and classify among the first-graders who have difficulty in school adjustment. Methods: Data from 1,010 children in the first grade of elementary school were used in the seventh to eighth surveys of the 2014-2015 Korean Children's Panel. The output variable of machine learning is the school adjustment type of the eighth survey data, and the type and number of each school adjustment were determined by latent profile analysis of its sub-variables. The input variables for machine learning were the children’s gender and school readiness of the seventh data. In addition, the children's difficulty in executive function, school preference, parents' interest in school life, teacher's teaching efficacy, and work stress included in the eighth data were selected as input variables. As a model for predicting children's school adjustment type, four machine learning algorithms are used. Results: Three types of school adjustment were identified through latent profile analysis, with the lowest level of school adjustment being 13.1% of all children. “Difficulty in executive function” was the predictor of the lowest level of school adjustment in all machine learning models. The model that best classifies children at the lowest level of school adjustment was the gradient-boosted decision tree model, with an accuracy of 94%, a sensitivity of 70%, a specificity of 97%, and an AUC of .89. Conclusion: When using latent profile analysis and machine learning model, the difficulty in adjusting to school in the first graders of elementary school was most predicted by difficulty of the executive function; however, school readiness was not. This study suggested that cooperation between early childhood education and primary education based on empirical analysis is necessary for successful primary connection.
关键词：Korean elementary school;first graders’ school adjustment;latent profile analysis;machine learning;discriminative model
