Exploring Transition Patterns and Key Predictors of dolescents’ Digital Usage and Competence: Applying XGBoost and Latent Transition Analysis

The purpose of this study is to classify the latent profiles and transition patterns of digital device usage and digital competence across elementary and middle school students, as well as to explore key predictors influencing these transitions. To this end, latent transition analysis, XGBoost, and SHAP(Shapley Additive Explanations), an explainable artificial intelligence algorithm, was applied to 1st and 2nd wave data of the elementary school students’ panel(n=3,464) and middle school students’ panel(n=3,556) of Seoul Educational Longitudinal Study of Students 2020(SELS 2020). The main findings are as follows. First, latent profiles were classified into three distinct groups across both educational levels, with differences observed in profile characteristics and proportions between panels. Second, while most students remained stable in their profiles, upward transitions in digital device usage and competence were more prevalent than downward transitions. Third, in the elementary school students’ panel, gender, guardians relationships, reading volume, academic burnout, among others were identified as key predictors. Fourth, in the middle school students’ panel, instructional quality in Korean language education, participation in extracurricular cultural activities, competence (cognitive, social, emotional), among others were derived as key predictors. Based on the findings, implications and recommendations for managing digital device usage and fostering digital competence in students are presented.