Digital device dependency among children has become a critical issue in the modern era, influencing cognitive, social, and health aspects. Excessive use of digital devices may lead to decreased concentration, academic performance, and social interaction. The identification of children's digital dependency levels has often relied on manual observation by parents or teachers, which tends to be subjective. Therefore, this study aims to develop a classification model for children's digital device dependency using the Learning Vector Quantization (LVQ) algorithm. The data were collected through a questionnaire distributed to 110 respondents, consisting of parents of elementary school students in Dumai City. The questionnaire contained 34 items measured using a five-point Likert scale (1–5). The data were processed using Python with supporting libraries such as NumPy, Pandas, Matplotlib, Scikit-learn, and Neupy. The experimental results showed that the LVQ algorithm successfully classified children's dependency levels into three categories low, moderate, and high with an accuracy of 87.5%, an average precision of 85.4%, and an average recall of 86.2%. The findings revealed that most children belong to the moderate dependency category, with an average score of 3.03. The main factors influencing digital dependency include usage duration, habits of using devices while eating or before sleeping, and decreased social interaction. The application of the LVQ algorithm proved effective in identifying children’s digital usage patterns and can serve as a foundation for developing early detection systems and promoting digital literacy policies within elementary education environments