A Study on the Development of a Customized Preventive Counseling Support Model for Students Based on Big Data

This study aims to develop a predictive model for student counseling support by identifying key individual factors encountered in university life, such as academic performance, financial conditions, psychological factors, and campus environment, as explanatory variables. Using logistic regression (logit model), the model was constructed to predict academic persistence and dropout, set as binary response variables, and validated for consistency through the support index. Additionally, the consistency of the dropout prediction model based on the logit model was cross-validated using a classification model with CART (Classification and Regression Tree) to determine which model or approach provided superior results. Modeling and comparing 2,000 simulated data points yielded similar results. All explanatory variables, except gender, were significant (p < .05), and, considering odds ratios, the number of leaves of absence (1.7818) and academic warnings (1.2921) had strong positive effects on dropout likelihood. Conversely, academic year, number of enrolled semesters, and GPA showed a negative effect on dropout probability. From a practical perspective for system implementation and operation, the logit model proved more efficient for constructing and applying the prediction model.