Machine Learning Classification of Insomnia Using Multidimensional Features and SMOTE

Novanza, Trional (Universitas Sriwijaya)
Ramadhan, Mgs. M. Luthfi (Universitas Sriwijaya)
Caryn, Femilia Hardina (Universitas Sriwijaya)
Passa, Rahma Satila (Universitas Sriwijaya)
Iqrom, Redho Aidil (Universitas Sriwijaya)
Seftianto, Ferlian (Universitas Sriwijaya)
Hardoni, Andre (Universitas Sriwijaya)
Baturohmah, Habi (Universitas Sriwijaya)

Jutisi: Jurnal Ilmiah Teknik Informatika dan Sistem Informasi
Vol 14, No 3: Desember 2025

Sleep disorders, especially insomnia, are common among adolescents and negatively affect health. Early detection is crucial for appropriate treatment. This study aims to classify insomnia severity in adolescents using a Machine Learning (ML) model and multidimensional features derived from 19 questionnaire instruments. The dataset consists of 95 adolescents aged 16–19 years, categorized into Insomnia, Subclinical Insomnia, and Control classes. The modeling process includes reducing multicollinearity, class balancing with SMOTE, and hyperparameter optimization using GridSearchCV and StratifiedKFold. Feature importance analysis was conducted using decision tree-based methods and permutation importance. The results show that SMOTE improves SVM performance from 0.690 to 0.793 and positively affects Random Forest. Logistic Regression performs best without SMOTE (accuracy 0.759), while XGBoost shows the lowest accuracy (0.614) even with SMOTE. A total of 11 features consistently contribute to all models. In conclusion, ML models, particularly SVM, are effective for classifying insomnia severity in adolescents.