TY - JOUR
T1 - Smooth support vector machine based on polynomial function for depression detection using electroencephalogram (EEG) signal
AU - Nikmah, Annisatul
AU - Purnami, Santi Wulan
AU - Andari, Shofi
AU - Maramis, Margarita M.
AU - Islamiyah, Wardah R.
AU - Zain, Jasni Muhammad
N1 - Publisher Copyright:
© 2024 AIP Publishing LLC.
PY - 2024/11/15
Y1 - 2024/11/15
N2 - Mental health is an important issue today as mental illness as a global health problem ranks fifth in the world. Depression is a major illness that affects many people around the world, and people suffering from depression often have a low level of awareness. It is still common to detect depression using clinical questionnaires. However, using questionnaires for large-scale surveys will consume large human and material resources. Therefore, scientists and researchers from around the world are working to find alternative and objective ways to detect mental depression, especially through EEG signal data. Several studies have shown that abnormal patterns in alpha waves in EEG signals are associated with depression. Still, beta, delta, theta, and gamma waves can also be used for depression detection. Before classification, EEG signal preprocessing is required by filtering using Finite Impulse Response (FIR). EEG signal data will be classified using one of the Machine Learning methods, namely Support Vector Machine (SVM), because, from some existing research, SVM provides superior performance compared to other methods. This research proposes Piecewise Polynomial Smooth Support Vector Machine (PPWSSVM) and Spline Smooth Support Vector Machine (Spline SSVM) for the classification method. This study found that, theoretically, the performance of the piecewise polynomial (PPWSSVM) function is better than the spline function. Classification using PPWSSVM with two channels, namely T3 and T4, provides the highest AUC value of 99.65% and 99.44%, respectively. While classification with one channel, namely T4, the highest AUC value uses Spline SSVM and SSVM.
AB - Mental health is an important issue today as mental illness as a global health problem ranks fifth in the world. Depression is a major illness that affects many people around the world, and people suffering from depression often have a low level of awareness. It is still common to detect depression using clinical questionnaires. However, using questionnaires for large-scale surveys will consume large human and material resources. Therefore, scientists and researchers from around the world are working to find alternative and objective ways to detect mental depression, especially through EEG signal data. Several studies have shown that abnormal patterns in alpha waves in EEG signals are associated with depression. Still, beta, delta, theta, and gamma waves can also be used for depression detection. Before classification, EEG signal preprocessing is required by filtering using Finite Impulse Response (FIR). EEG signal data will be classified using one of the Machine Learning methods, namely Support Vector Machine (SVM), because, from some existing research, SVM provides superior performance compared to other methods. This research proposes Piecewise Polynomial Smooth Support Vector Machine (PPWSSVM) and Spline Smooth Support Vector Machine (Spline SSVM) for the classification method. This study found that, theoretically, the performance of the piecewise polynomial (PPWSSVM) function is better than the spline function. Classification using PPWSSVM with two channels, namely T3 and T4, provides the highest AUC value of 99.65% and 99.44%, respectively. While classification with one channel, namely T4, the highest AUC value uses Spline SSVM and SSVM.
UR - http://www.scopus.com/inward/record.url?scp=85210223629&partnerID=8YFLogxK
U2 - 10.1063/5.0239089
DO - 10.1063/5.0239089
M3 - Conference article
AN - SCOPUS:85210223629
SN - 0094-243X
VL - 3201
JO - AIP Conference Proceedings
JF - AIP Conference Proceedings
IS - 1
M1 - 060014
T2 - 9th SEAMS-UGM International Conference on Mathematics and its Applications 2023: Integrating Mathematics with Artificial Intelligence to Broaden its Applicability through Industrial Collaborations
Y2 - 25 July 2023 through 28 July 2023
ER -