Orthopedic diseases significantly affect the musculoskeletal system, reducing patients’ functional capacity and quality of life. Accurate and early classification of such conditions is therefore critical for effective clinical decision-making. This study proposes a machine learning-based framework for orthopedic disease classification using biomechanical features, with a particular emphasis on handling class imbalance. A set of classification algorithms, including Random Forest, Support Vector Machine, Naive Bayes, Logistic Regression, XGBoost, LightGBM, and a Soft Voting Ensemble (XGBoost + LightGBM), were evaluated on a dataset of 310 patients using 10-fold cross-validation. The impact of the Synthetic Minority Over-sampling Technique (SMOTE) was systematically analyzed by comparing model performance with and without its application. Evaluation metrics included Accuracy, Precision, Recall, F1-score, and macro-average ROC-AUC. Results indicate that addressing class imbalance significantly improves model performance, particularly in terms of ROC-AUC. Among the tested methods, Logistic Regression demonstrated the most stable and competitive results. The best performance was achieved by Logistic Regression with SMOTE, yielding 87% accuracy and a macro-average ROC-AUC of 0.96. These findings highlight the importance of imbalance-aware modeling strategies in orthopedic disease classification.
Özbalcı et al. (Tue,) studied this question.