Harper, M (2023) Comparison of subjective and physiological stress levels in home and office work environments. Doctoral thesis, Liverpool John Moores University.

Preview

Text Comparison of subjective and physiological stress levels in home and office work environments.pdf - Published Version Available under License Creative Commons Attribution Non-commercial. Download (1MB) | Preview

Abstract

Work stress is a major problem to individuals and society, with prolonged periods of stress often leading to health issues and reduced productivity. COVID-19 has increased the incidence of individuals working in a mixture of home and office-based environments, with each location presenting its own stressors. Identification of stress levels in each environment will allow individuals to better plan how to mitigate stress and boost productivity. In this project, differences in stress levels are predicted in each work environment from individuals’ physiological responses and subjectively reported stress and productivity. Initial work on the project focused upon development of a system for the detection of dementia-related difficulties through the wearable-based tracking of physiological indicators. As such, a review of the available commercial and laboratory devices available for tracking physiological indicators of dementia-related difficulties was conducted. Furthermore, no publicly available physiological dataset for predicting difficulties in dementia currently exists. However, a review of the methods for collecting such a dataset and the impact of COVID-19 found that it is impractical and potentially unethical to conduct an experiment with people with dementia during the pandemic. As such, a pivot in research was necessitated. Comparing the stress levels of individuals working in home and office environments was selected. A data collection experiment was then performed with 13 academics working in combinations of home and office environments. Descriptive statistical features were then extracted from both the physiological and questionnaire data, with the relationships between attributes and features calculated using various advanced data analytics and statistical approaches. The resultant correlation coefficients and statistical summaries of stress were used to evaluate relationships between stress and work environment at different times of day, different days of the week, and while performing different activities. A bagged tree machine learning model was trained over the data, achieving 99.3% accuracy when evaluated using 10-fold cross validation. When tested on the purely unseen instances it achieved 56% accuracy corresponding to inter-class stress classification, however a testing accuracy of 73.7% was achieved using principal component analysis for dimensionality reduction and the dataset is balanced using Synthetic Minority Oversampling Technique.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Machine learning; Wearable computing; Work stress
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QP Physiology R Medicine > RA Public aspects of medicine > RA0421 Public health. Hygiene. Preventive Medicine
Divisions:	Computer Science & Mathematics
SWORD Depositor:	A Symplectic
Date Deposited:	29 Sep 2023 14:48
Last Modified:	29 Sep 2023 14:49
DOI or ID number:	10.24377/LJMU.t.00021630
Supervisors:	Khan, W, Al-Jumeily OBE, D, Ghali, F and Hussain, A
URI:	https://researchonline.ljmu.ac.uk/id/eprint/21630

View Item