<![CDATA[Sleep disorders are increasingly prevalent health issues that significantly affect individual’s quality of life. Timely detection and accurate classification of these disorders are essential for proper diagnosis and effective clinical intervention. However, a major challenge in classifying sleep disorders lies in the imbalance of data distribution—where majority classes have substantially more data than minority ones. This imbalance often leads to predictive models that favor the dominant class, thereby reducing overall classification accuracy. This study focuses on enhancing sleep disorder classification performance on imbalanced datasets by applying the Synthetic Minority Over-sampling Technique (SMOTE) to balance the data. It also evaluates the effectiveness of various machine learning algorithms in identifying sleep disorders. The algorithms analyzed include Random Forest (RF), Neural Network (NN), Naive Bayes (NB), K-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Logistic Regression (LR), tested both before and after applying SMOTE. Model performance was assessed using accuracy, precision, recall, and F1-score to ensure a comprehensive evaluation. The findings indicate that SMOTE consistently boosts the performance of all tested models. Among them, the Neural Network combined with SMOTE achieved the highest performance, with an accuracy of 92.00%, precision of 91.88%, recall of 92.00%, and an F1-score of 91.91%. Additionally, the Random Forest model with SMOTE produced the highest F1-score at 93.18%, demonstrating strong performance stability. These results highlight the effectiveness of integrating oversampling techniques like SMOTE with machine learning models to address class imbalance, leading to more accurate and reliable classification outcomes. The study offers valuable insights for developing AI-based medical decision support systems focused on sleep disorder diagnosis.]]>
Copyrights © 2025
