Zahra, Fadhila Fathin (2026) Integrasi Fitur Periodik Dan Periodik Berbasis FOOOF Pada Bagging SVM Untuk Klasifikasi Tingkat Depresi Pasca Terapi Mindfulness Dengan Interpretasi SHAP. Masters thesis, Institut Teknologi Sepuluh Nopember.
|
Text
Tesis_Fadhila Fathin Zahra_6003242008.pdf - Accepted Version Restricted to Repository staff only Download (29MB) | Request a copy |
Abstract
Depresi merupakan salah satu gangguan kesehatan mental yang banyak terjadi di seluruh dunia dengan tingkat keparahan yang beragam. Salah satu metode non-farmakologis yang terbukti efektif dalam mengurangi gejala depresi adalah Mindfulness-Based Cognitive Therapy (MBCT). Namun, evaluasi efektivitas terapi tersebut masih memerlukan pendekatan yang lebih objektif berbasis biomarker otak. Penelitian ini mengusulkan kerangka klasifikasi tingkat depresi berbasis sinyal Electroencephalogram (EEG) untuk mengevaluasi kondisi depresi sebelum dan setelah terapi mindfulness. Fitur EEG diekstraksi menggunakan metode Fitting Oscillations and One-Over-F (FOOOF), yang menghasilkan komponen periodik dan aperiodik, termasuk parameter offset dan eksponen. Data kemudian dikelompokkan ke dalam tiga kategori, yaitu normal, depresi tingkat 1, dan depresi tingkat 2 berdasarkan hasil pelabelan RFCM. Fitur-fitur yang telah melalui proses seleksi digunakan sebagai masukan pada model Bagging Multiclass Support Vector Machine (SVM). Hasil penelitian menunjukkan bahwa model yang diusulkan mampu mengklasifikasikan tingkat depresi dengan performa yang sangat baik, yaitu akurasi 95,2 persen, presisi 95,4 persen, recall 95,3 persen, F1-score 95,3 persen, dan AUC 0,971. Untuk meningkatkan interpretabilitas model yang bersifat black box, digunakan metode SHapley Additive exPlanations (SHAP). Hasil interpretasi menunjukkan bahwa fitur aperiodik, khususnya offset dan eksponen, merupakan faktor yang paling berpengaruh dalam proses klasifikasi, diikuti oleh amplitudo gelombang alpha pada area frontal otak. Temuan ini sejalan dengan teori Excitation-Inhibition Balance pada jaringan saraf serta peran area prefrontal dalam regulasi emosi. Secara keseluruhan, kombinasi metode FOOOF, Bagging Multiclass SVM, dan SHAP berhasil menghasilkan sistem klasifikasi tingkat depresi berbasis EEG yang akurat, objektif, dan mudah diinterpretasikan. Hasil penelitian ini menunjukkan bahwa parameter aperiodik dan amplitudo alpha berpotensi menjadi biomarker neurofisiologis untuk identifikasi tingkat depresi serta evaluasi efektivitas terapi mindfulness secara lebih objektif.
====================================================================================================================================
Depression is one of the most common mental health disorders worldwide, with varying levels of severity. One non-pharmacological method proven to be effective in reducing depressive symptoms is Mindfulness-Based Cognitive Therapy (MBCT). However, evaluating the effectiveness of this therapy still requires a more objective approach based on brain biomarkers. This study proposes a depression level classification framework based on electroencephalogram (EEG) signals to evaluate depression conditions before and after mindfulness therapy. EEG features were extracted using the Fitting Oscillations and One-Over-F (FOOOF) method, which generates periodic and aperiodic components, including offset and exponent parameters. The data were then grouped into three categories normal, depression level 1, and depression level 2 based on RFCM labeling results. The selected features were used as inputs for the Bagging Multiclass Support Vector Machine (SVM) model. The results showed that the proposed model achieved excellent performance in classifying depression levels, with an accuracy of 95.2 percent, precision of 95.4 percent, recall of 95.3 percent, F1-score of 95.3 percent, and an AUC of 0.971. To enhance the interpretability of the black-box model, the SHapley Additive exPlanations (SHAP) method was employed. The interpretation results indicate that aperiodic features, particularly offset and exponent, are the most influential factors in the classification process, followed by alpha wave amplitude in the frontal brain area. These findings align with the excitation-inhibition balance theory in neural networks and the role of the prefrontal cortex in emotion regulation. Overall, the combination of FOOOF, Bagging Multiclass SVM, and SHAP successfully produced an EEG-based depression level classification system that is accurate, objective, and interpretable. The results of this study suggest that aperiodic parameters and alpha amplitude have the potential to serve as neurophysiological biomarkers for identifying depression levels and evaluating the effectiveness of mindfulness therapy more objectively.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | Depresi, EEG, FOOOF, Fitur Aperiodik, Bagging SVM, SHAP, Mindfulness, Depression, Mindfulness, EEG, FOOOF, Aperiodic Features, Bagging SVM, SHAP |
| Subjects: | Q Science Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. |
| Divisions: | Faculty of Mathematics and Science > Statistics > 49101-(S2) Master Thesis |
| Depositing User: | Fadhila Fathin Zahra |
| Date Deposited: | 30 Jul 2026 08:13 |
| Last Modified: | 30 Jul 2026 08:13 |
| URI: | http://repository.its.ac.id/id/eprint/139684 |
Actions (login required)
![]() |
View Item |
