Preliminary Analysis of Human Error Prediction Model by Using Biological Information

Abstract

Increasing in aging population forced the society to act more than their limit. For instance, an action such as driving, where we need our mental concentration at most, could lead to serious accident from a simple mistake because of overwork. Therefore, it is crucial to prevent the accident. Many researchers focus on biological information to predict the error because human error always related to a person’s cognitive condition such as stress and discomfort. However, existing studies on the human error prediction model have not conducted a detailed analysis, and also have not considered individual differences. Therefore, the purpose of this study is to analyze the biological information immediately before and after the occurrence of human error in order to construct a prediction model for human error considering individual differences. In this study, we developed the Stroop task to be used as the mental workload and measured the subjects’ biological information. As a result, we proposed 10 [s] as the time intervals for before and after the consecutive of the occurrence of the human errors for better analysis. Besides, the biological information measured from all subjects suggested that pNN10 can be considered as the predictive indicator for human error occurrence. However, other biological information also expressed vary results where our next step needs to consider the individual differences by increasing the sample size. In addition, the logistic regression will be considered for machine learning to be used for the human error prediction model construction.