This study investigates the prediction of students’ adaptability levels in online education using machine learning, with a focus on decision tree–based algorithms. Guided by the objectives of dataset collection, framework design, model optimization, and evaluation, the study utilized secondary datasets sourced from existing research and an open university in the United States. Data preprocessing involved handling missing values, encoding categorical variables, scaling, and addressing class imbalance with SMOTE, followed by feature selection using RFECV. Four algorithms, namely C4.5-like Decision Tree, RepTree-like Decision Tree, RandomTree, and Random Forest, were optimized with Bayesian tuning and evaluated using accuracy, precision, recall, and F1-score. The results revealed RandomTree as the best-performing model (accuracy = 78.84%, F1 = 0.7892), a key finding that challenges the common superiority of Random Forest in related research. A generalizability assessment through nested cross-validation confirmed RandomTree’s robustness (accuracy = 84.40% ± 0.0177 (mean/standard deviation)). Feature importance analysis highlighted age, financial condition, internet type, and location as critical determinants of adaptability. The findings emphasize the interplay of demographic, socio-economic, and technological factors in online learning adaptability, offering both theoretical contributions to educational data mining and practical tools for institutions to monitor and support students.

Online Student adaptability Students Adaptability Prediction Model Interpretable Students’ Adaptability Model Decision Tree Based Students Adaptability Model