<![CDATA[Background: Educational data mining has become increasingly important for understanding student performance and learning patterns; however, challenges in data preprocessing, feature selection, interpretability, and the limited integration of predictive modeling with clustering remain unresolved. Objective: This study develops an integrated machine-learning workflow that combines feature engineering, predictive modeling, and student segmentation for educational data analysis. Methods: The analytical dataset contains 343 student records and 87 numeric features after data-type conversion, with Adaptability as the prediction target. The workflow combines data-quality control, feature engineering, ANOVA, mutual information, and recursive feature elimination (RFE) for feature selection; comparative classification; stratified cross-validation; and K-means clustering with principal component analysis (PCA) for student segmentation. Results: Logistic Regression produced the strongest reported discrimination (Accuracy = 0.9855; Precision = 1.0000; Recall = 0.9697; F1 = 0.9846; MCC = 0.9713; Kappa = 0.9709; AUC = 0.997). AUC values for the remaining models ranged from 0.958 (KNN) to 0.996 (Gradient Boosting). Feature rankings are interpreted as predictive associations rather than causal effects, and clustering provides descriptive student subgroups. Conclusion: The study provides a traceable, dataset-specific workflow linking preprocessing, feature construction and selection, predictive benchmarking, and descriptive student segmentation. Generalization beyond the supplied dataset requires external validation, and the clustering results should not be interpreted as evidence of intervention effects.]]>
