Construction of a Predictive Model for Theory and Practical Performance in '

Background: Materials and Methods: Machine learning techniques were applied to develop predictive models using data from undergraduate nursing students who enrolled the 2019 and 2020 cohorts. Demographic and learning behavior variables were collected and divided into training and testing sets (7:3 ratio). Least absolute shrinkage and selection operator regression was used for feature selection, and logistic regression identified independent risk factors. Five algorithms [Logit, RF, XGBoost, Artificial Neural Network, and Support Vector Machine (SVM)] were used to construct models. Model performance was evaluated using area under the curve (AUC), specificity, and sensitivity. Results: Data from 527 students were analyzed; 59 (11.19%) failed the theoretical exam, and 116 (22.01%) failed the practical exam. Predictors for practical exam included height, comprehensive assessment scores, periodic exam scores, comprehensive practical training scores, and practice frequency. Predictors for the theoretical exam included gender, research participation, academic background, height, periodic exam, and practice frequency. SVM showed the best performance for practical exams prediction (AUC = 0.737), while XGB performed best for theoretical exams (AUC = 0.833). No statistically significant differences in AUC were found between most models ( Conclusions: Height, assessment scores, training engagement, exam results, gender, and practice frequency significantly influenced nursing students' academic outcomes. These findings support personalized teaching strategies in foundational nursing education.