Fathiyaturrahmi, Laila (2025) Semi-supervised Clustering dengan Fuzzy constraint untuk Pengelompokan Mental state Berdasarkan Rekaman Electroencephalogram. Masters thesis, Institut Teknologi Sepuluh Nopember.
Text
6003231011 -Master_thesis.pdf Restricted to Repository staff only until 1 April 2027. Download (5MB) | Request a copy |
Abstract
Electroencephalography merupakan metode merekam aktivitas elektrik otak yang dilakukan pada permukaan kulit kepala menghasilkan sinyal Electroencephalo-gram (EEG). Data EEG dimanfaatkan untuk mendiagnosis mental state. Peme-riksaan mental state melibatkan kondisi netral, rileks dan konsentrasi. Penge-lompokan mental state berdasarkan rekaman EEG dapat dilakukan dengan fuzzy Hierarchical Semi-supervised (HSS) clustering yang memanfaatkan fuzzy cons-traint untuk menentukan titik potong (η-cut) optimal. Penelitian menunjukkan nilai koefisien silhoette pada data real dengan channel oksipital kiri (O1) dan oksipital kanan (O2) menggunakan HSS dengan fuzzy constraint berturut-turut 0,57 dan 0,58 dibandingkan dengan HSS dengan hard constraint yaitu 0,08 dan 0,18. Hal tersebut menunjukkan bahwa fuzzy HSS mampu mempertahankan homogenistas dalam cluster dan heterogenitas antar cluster. Cluster yang dihasilkan dalam fuzzy HSS dapat dinyatakan mampu menyesuaikan dengan kondisi alami data meskipun telah diberikan constraint. Hal tersebut terlihat dari NMI dan purity pada fuzzy HSS yang lebih yaitu 0,14 dan 0,13 pada channel O1 dan O2 disertai dengan nilai koefisien silhouette yang relatif baik dibandingkan Hard HSS yang memiliki nilai NMI dan purity lebih besar yaitu 0,50 dan 0,46, namun memiliki anggota cluster yang sangat kecil atau relatif terpisah. Hasil tersebut didukung oleh pengelompokan mental state menggunakan data publik, dimana nilai koefisien silhoette pada seluruh channel yang digunakan menggunakan fuzzy HSS lebih besar dibandingkan Hard HSS.
================================================================================================================================
Electroencephalography is a method for recording the brain's electrical activity on the scalp surface, producing Electroencephalogram (EEG) signals. EEG data is utilized to diagnose mental states, which involve conditions such as neutral, relaxed, and focused states. Mental state classification based on EEG recordings can be performed using fuzzy Hierarchical Semi-supervised (HSS) clustering, which leverages fuzzy constraints to determine the optimal cut point (η-cut). The study demonstrates that the silhouette coefficient values for real data using channels O1 and O2 with HSS and fuzzy constraints are 0.57 and 0.58, respectively, compared to HSS with hard constraints, which achieve 0.08 and 0.18. These results indicate that fuzzy HSS effectively maintains intra-cluster homogeneity and inter-cluster heterogeneity. The clusters produced by fuzzy HSS are shown to adapt to the natural characteristics of the data, even when constraints are applied. This is evident from the NMI and purity values in fuzzy HSS, which are 0.14 and 0.13 for channels O1 and O2, respectively, accompanied by relatively good silhouette coefficient values compared to hard HSS, which has higher NMI and purity values (0.50 and 0.46) but smaller or more separated cluster members. These findings are supported by mental state classification using public data, where the silhouette coefficient values across all channels are consistently higher with fuzzy HSS compared to hard HSS.
Item Type: | Thesis (Masters) |
---|---|
Uncontrolled Keywords: | Electroencephalogram, Semi-supervised Clustering, Fuzzy constraint |
Subjects: | Q Science > QA Mathematics > QA278.55 Cluster analysis Q Science > QA Mathematics > QA278 Cluster Analysis. Multivariate analysis. Correspondence analysis (Statistics) Q Science > QP Physiology > Q376.5 Electroencephalography (EEG) |
Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis |
Depositing User: | Laila Fathiyaturrahmi |
Date Deposited: | 03 Feb 2025 01:46 |
Last Modified: | 03 Feb 2025 01:46 |
URI: | http://repository.its.ac.id/id/eprint/117837 |
Actions (login required)
View Item |