A Study of Implementing Artificial Neural Networks and Cluster Analysis to Distinguish Fatigue Type and Level in Graduate Students

Abstract

There are many causes of fatigue and the prevalence rate and key factors of fatigue differ according to population groups. Masters students in Taiwan are a high risk group for fatigue, and their lifestyle data was collected as the subject of the present study. Physiological parameters are measured using mobile devices, and the checklist individual strength (CIS) questionnaire and fatigue type checklist are utilized to explore the prevalence rate and type of fatigue in masters students. Cluster analysis was used to establish fatigue levels, and Pearson's correlation analysis was used to explore the correlation between different fatigue types and fatigue level. The results obtained from the CIS questionnaire showed a fatigue prevalence rate of 50%, with a Cronbach's alpha value of 0.885, indicating good internal consistency. Fatigue type was established using the fatigue type checklist and a neural network. Masters students who are fatigued with an active sympathetic nervous system accounted for 28.75% of the subjects, while 21.75% of the subjects were fatigued with an active parasympathetic nervous system, where the key factors are the number of exercise days and the number of steps taken. Cluster analysis was then used to separate the degree of fatigue into four levels. Fatigue scores between 111 and 140 are classified as extremely fatigued; between 77 and 110 as generally fatigued; between 48 and 76 as borderline fatigued; and between 20 and 47 as removed from fatigue. Different fatigue levels were found to have different key factors, and the results of the present study can help provide differentiated solutions for different levels of fatigue.