Prediksi Tingkat Kelelahan pada Petugas Pemadam Kebakaran Berdasarkan Pola Tidur, Sistem Shift, dan Indikator Fisiologis

Yogatama, Rozaan Naufanuha (2026) Prediksi Tingkat Kelelahan pada Petugas Pemadam Kebakaran Berdasarkan Pola Tidur, Sistem Shift, dan Indikator Fisiologis. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5010221155_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Petugas pemadam kebakaran menghadapi risiko kelelahan kerja yang tinggi akibat sistem shift, tingginya frekuensi alarm darurat, serta tuntutan transisi fisiologis yang mendadak dari kondisi istirahat menuju aktivitas berintensitas maksimal. Kondisi tersebut memicu akumulasi kelelahan laten, yaitu penurunan fungsi tubuh yang terjadi secara tersembunyi tanpa gejala fisik yang tampak, namun terbukti berkontribusi terhadap 40–50% kematian petugas akibat Sudden Cardiac Death setiap tahunnya. Meskipun berbagai penelitian telah mengkaji hubungan antara kelelahan, pola tidur, dan sistem shift pada petugas pemadam kebakaran, masih terdapat empat kesenjangan penelitian yang belum terjawab, yaitu dominasi instrumen subjektif berbasis kuesioner, belum adanya model prediksi berbasis parameter fisiologis yang objektif, belum terintegrasinya parameter Heart Rate Variability (HRV) dengan data pola tidur dan shift kerja dalam satu kerangka prediksi, serta belum adanya penelitian yang memosisikan Psychomotor Vigilance Test (PVT) sebagai ground truth objektif untuk memvalidasi status kelelahan secara real-time. Penelitian ini bertujuan membangun model prediksi tingkat kelelahan petugas pemadam kebakaran menggunakan tiga algoritma supervised learning, yaitu Logistic Regression, Decision Tree, dan Random Forest, berdasarkan tiga variabel masukan, yaitu nilai RMSSD (Root Mean Square of Successive Differences) yang diekstraksi dari rekaman HRV short-term selama lima menit menggunakan heart rate monitor armband, durasi tidur yang dicatat melalui sleep diary, dan jenis shift kerja. Status kelelahan diklasifikasikan menjadi tiga kategori berdasarkan hasil PVT dengan 30 stimulus dan ambang batas (threshold) 355 ms yang mengacu pada kerangka kerja Basner dan Rubinstein (2011) yang diadaptasi secara proporsional, yaitu Fit apabila menunjukkan kurang dari 6 lapses (label: 0), Fit with Note apabila menunjukkan 6–10 lapses (label: 1), dan Unfit apabila menunjukkan lebih dari 10 lapses (label: 2) sebagai ground truth yang objektif dan terukur. Penelitian dilakukan menggunakan desain cross-sectional pada personel aktif Dinas Pemadam Kebakaran Pos Keputih Surabaya dan Pos Mulyorejo. Evaluasi model difokuskan pada metrik Recall sebagai parameter utama untuk meminimalkan risiko False Negative dalam konteks keselamatan kerja, serta dilengkapi dengan pengukuran akurasi, presisi, dan F1-Score. Model terbaik yang dihasilkan diharapkan dapat menjadi instrumen deteksi dini kelelahan yang objektif, akurat, dan aplikatif sebagai dasar pengembangan sistem manajemen keselamatan kerja berbasis data fisiologis di lingkungan dinas pemadam kebakaran.
=================================================================================================================================
Firefighters face a high risk of occupational fatigue due to shift work, the high frequency of emergency alarms, and the demands of sudden physiological transitions from a resting state to maximum-intensity activity. These conditions contribute to the accumulation of latent fatigue, a decline in physiological function that occurs without visible physical symptoms but has been shown to contribute to 40–50% of firefighter deaths caused by Sudden Cardiac Death each year. Although numerous studies have investigated the relationship between fatigue, sleep patterns, and shift systems among firefighters, four major research gaps remain: the predominance of subjective questionnaire-based assessment methods, the absence of predictive models based on objective physiological parameters, the lack of integration of Heart Rate Variability (HRV) parameters with sleep pattern and work shift data within a single predictive framework, and the absence of studies that position the Psychomotor Vigilance Test (PVT) as an objective ground truth for real-time fatigue validation. This study aims to develop a firefighter fatigue prediction model using three supervised learning algorithms, namely Logistic Regression, Decision Tree, and Random Forest, based on three input variables: RMSSD (Root Mean Square of Successive Differences) values extracted from five-minute short-term HRV recordings using a heart rate monitor armband, sleep duration recorded through a sleep diary, and work shift type. Fatigue status is classified into three categories based on the results of a 30-stimulus PVT using a threshold of 355 ms, proportionally adapted from the framework proposed by Basner and Rubinstein (2011): Fit for fewer than six lapses (label: 0), Fit with Note for six to ten lapses (label: 1), and Unfit for more than ten lapses (label: 2), serving as an objective and measurable ground truth. The study employed a cross-sectional design involving active personnel from the Surabaya Fire Department at the Keputih and Mulyorejo stations. Model evaluation primarily focuses on Recall to minimize the risk of False Negatives in an occupational safety context, complemented by accuracy, precision, and F1-Score. The best-performing model is expected to provide an objective, accurate, and practical early fatigue detection tool, forming the basis for the development of a physiological data-driven occupational safety management system within fire department operations.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Deteksi dini, Heart Rate Variability, kelelahan kerja, machine learning, pemadam kebakaran, pola tidur, Psychomotor Vigilance Test, RMSSD, shift kerja, supervised learning. Early detection, Firefighters, Heart Rate variability, Machine learning, Psychomotor Vigilance Test, RMSSD, Sleep patterns, Supervised learning, Work-related fatigue, Work shifts.
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Industrial Engineering > 26201-(S1) Undergraduate Thesis
Depositing User:	Rozaan Naufanuha Yogatama
Date Deposited:	13 Jul 2026 09:03
Last Modified:	13 Jul 2026 09:03
URI:	http://repository.its.ac.id/id/eprint/134810

Actions (login required)

View Item