Analisis Cognitive Workload Berdasarkan Parameter Fisiologis Manusia Menggunakan Random Forest Classifier Dalam Studi Kasus Pengembangan Produk Dan Perencanaan Produksi

Harmayanti, Afifah and Ir. Ishardita Pambudi Tama, ST., MT., Ph.D. and Prof. Dr. Ir. Femiana Gapsari, ST., MT. (2024) Analisis Cognitive Workload Berdasarkan Parameter Fisiologis Manusia Menggunakan Random Forest Classifier Dalam Studi Kasus Pengembangan Produk Dan Perencanaan Produksi. Magister thesis, Universitas Brawijaya.

Abstract

Salah satu faktor penengah antara human factors dan kualitas proses produksi adalah fatigue pada manusia, yang dapat ditinjau dari aspek ergonomi fisik dan ergonomi kognitif. Smart manufacturing system, telah memunculkan lebih banyak pekerjaan kognitif untuk para pekerja industri manufaktur. Sehingga, analisis cognitive workload dibutuhkan untuk mengevaluasi tingkat dan dimensi workload yang dialami oleh pekerja. Perubahan HR, low frequency power (LF), total frequency power (TP) dan rasio low frequency terhadap high frequency power (rasio LF/HF), dipengaruhi oleh perubahan aktivitas mental. Sehingga parameter – parameter tersebut dapat digunakan untuk menganalisis cognitive workload pada individu yang sedang melakukan aktivitas kognitif. Penelitian kali ini, analisis cognitive workload dilakukan terhadap 30 partisipan mahasiswa Teknik Mesin dan Teknik Industri, menggunakan studi kasus pemecahan masalah kognitif, yang dirancang menyerupai studi kasus di bidang pengembangan produk dan perencanaan produksi. Pengukuran HR, TP, LF dan rasio LF/HF dilakukan menggunakan Polar H10 chest strap. Keempat parameter distandarisasi menggunakan effect size untuk melihat besarnya perubahan indikator tersebut sebelum dan ketika mengerjakan aktivitas. Kemudian hasil perhitungan effect size dimasukkan ke dalam permodelan random forest classifier, untuk mendapatkan klasifikasi level dan domain cognitive workload. Validasi hasil klasifikasi menggunakan kuesioner NASA – TLX. Hasil penelitian menunjukkan bahwa permodelan random forest classifier mampu mencapai akurasi sebesar 90,9% untuk klasifikasi level dan 81,8% untuk klasifikasi domain cognitive workload. Sementara analisis features importance menunjukkan bahwa perubahan TP HRV menjadi indikator yang paling berpengaruh pada permodelan level cognitive workload. Sedangkan pada analisis domain cognitive workload, perubahan rasio LF/HF yang paling berpengaruh terhadap permodelan tersebut.

English Abstract

One of the intermediate factors between human factors and production quality is human fatigue. Fatigue can be evaluated using both physical and cognitive ergonomic approach. Smart manufacturing system has open up a range of cognitive based job tasks in the manufacturing industry. Therefore, cognitive workload analysis is needed to analyse the level and domain of workload a worker’s experiencing. In prior studies, heart rate (HR), HRV low frequency power (LF), HRV total frequency power (TP) and ratio of HRV low and high frequency power (Ratio LF/HF), were found to be correlated to mental state changes. Thus, those parameters could be used to analyse cognitive workload during cognitive engaged activities. In this thesis research, cognitive workload analysis was carried out to 30 participants form Mechanical and Industrial Engineering students, using a designed problem – solving questions, based on cognition concept and cases in the product development and production planning fields. The participants HR, LF, TP and LF/HF Ratio were recorded using a Polar H10 chest strap before and during the activity. All physiological indicators were then standardized using effect size approach to measure the magnitude of physiological changes due to the induced cognitive stimulation. After the four indicators had been standardized, they were input to the random forest classifier model to classify the level and domain of cognitive workload experienced by the participants. Results of NASA – TLX questionnaire from each participant, were used to validate the classification made by the model. The result, an accuracy of 90.9% was obtained for the level classification and 81,8% was obtained for the domain classification. As for the features importance analysis, HRV TP changes was the most important feature to the level classification model. Meanwhile, the LF/HF Ratio changes was found to be the most important feature to the domain classification model.

Item Type:	Thesis (Magister)
Identification Number:	0424070013
Uncontrolled Keywords:	cognitive workload, manufaktur, random forest classifier, HR, HRV-cognitive workload, manufacturing, random forest classifier, HR, HRV
Divisions:	S2/S3 > Magister Teknik Mesin, Fakultas Teknik
Depositing User:	Sugeng Moelyono
Date Deposited:	02 May 2024 07:15
Last Modified:	02 May 2024 07:15
URI:	http://repository.ub.ac.id/id/eprint/218806

Text (DALAM MASA EMBARGO) Afifah Harmayanti.pdf Restricted to Registered users only Download (5MB)

Actions (login required)

View Item