KLASIFIKASI JENIS GANGGUAN TIDUR MENGGUNAKAN METODE LONG SHORT TERM MEMORY (LSTM)

SYACHPUTRA, BAGUS INDRA and Arsalan, Osvari and Kurniati, Junia (2025) KLASIFIKASI JENIS GANGGUAN TIDUR MENGGUNAKAN METODE LONG SHORT TERM MEMORY (LSTM). Undergraduate thesis, Sriwijaya University.

[thumbnail of RAMA_55201 _09021382025116_Cover.pdf] Image
RAMA_55201 _09021382025116_Cover.pdf - Cover Image
Available under License Creative Commons Public Domain Dedication.
Download (1MB)
[thumbnail of RAMA_55201 _09021382025116.pdf] Text
RAMA_55201 _09021382025116.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.
Download (3MB)
[thumbnail of RAMA_55201 _09021382025116_TURNITIN.pdf] Text
RAMA_55201 _09021382025116_TURNITIN.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.
Download (10MB)
[thumbnail of RAMA_55201 _09021382025116_0028068806_0026068907_01_front_ref.pdf] Text
RAMA_55201 _09021382025116_0028068806_0026068907_01_front_ref.pdf - Accepted Version
Available under License Creative Commons Public Domain Dedication.
Download (1MB)
[thumbnail of RAMA_55201 _09021382025116_0028068806_0026068907_02.pdf] Text
RAMA_55201 _09021382025116_0028068806_0026068907_02.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.
Download (674kB)
[thumbnail of RAMA_55201 _09021382025116_0028068806_0026068907_03.pdf] Text
RAMA_55201 _09021382025116_0028068806_0026068907_03.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.
Download (269kB)
[thumbnail of RAMA_55201 _09021382025116_0028068806_0026068907_04.pdf] Text
RAMA_55201 _09021382025116_0028068806_0026068907_04.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.
Download (1MB)
[thumbnail of RAMA_55201 _09021382025116_0028068806_0026068907_05.pdf] Text
RAMA_55201 _09021382025116_0028068806_0026068907_05.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.
Download (398kB)
[thumbnail of RAMA_55201 _09021382025116_0028068806_0026068907_06_ref.pdf] Text
RAMA_55201 _09021382025116_0028068806_0026068907_06_ref.pdf - Bibliography
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.
Download (134kB)
[thumbnail of RAMA_55201 _09021382025116_0028068806_0026068907_07_lamp.pdf] Text
RAMA_55201 _09021382025116_0028068806_0026068907_07_lamp.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.
Download (175kB)

Abstract

Sleep quality plays a critical role in maintaining physical, mental, and emotional wellbeing. Early detection of sleep disorders such as insomnia and sleep apnea is essential for enabling more effective clinical interventions. This study investigates the optimization of sleep disorder classification using Long Short-Term Memory (LSTM) networks based on time-series data incorporating 13 physiological and behavioral features including sleep duration, heart rate, and stress levels. The preprocessing pipeline included data normalization, categorical feature encoding, and decomposition of blood pressure components to ensure input consistency. We systematically evaluated the LSTM architecture through comprehensive hyperparameter tuning, examining variations in: the number of LSTM units, dropout rates, dense layer configurations, training epochs, and batch sizes. The optimal model configuration achieved 92% classification accuracy with 16 LSTM units, a dropout rate of 0.2, a dense layer of 32 units, and a batch size of 8. This configuration yielded precision of 91.99%, recall of 92%, and F1-score of 91.88%. Model convergence occurred at epoch 16 without evidence of overfitting. Notably, increasing LSTM units to 32 or 64 resulted in reduced accuracy (89.33%), suggesting increased susceptibility to overfitting with larger architectures. These findings demonstrate the effectiveness of LSTM networks in identifying complex temporal patterns associated with sleep disorders while emphasizing the importance of careful hyperparameter optimization for classification performance. The proposed approach shows promise for developing more accurate and efficient digital tools for sleep disorder detection and assessment.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Klasifikasi, Gangguan Tidur, Long Short Term Memory
Subjects: T Technology > T Technology (General) > T1-995 Technology (General)
Divisions: 09-Faculty of Computer Science > 55201-Informatics (S1)
Depositing User: Bagus Indra Syachputra
Date Deposited: 17 Sep 2025 01:33
Last Modified: 17 Sep 2025 01:33
URI: http://repository.unsri.ac.id/id/eprint/184020

Actions (login required)

View Item View Item