Modeling noise annoyance using Bayesian networks: A case study of an industrial factory

BackgroundNoise annoyance is a detrimental cognitive response elicited by noise exposure. Precisely identifying primary risk factors plays a vital role in effectively preventing and managing this issue.ObjectiveThis study aimed to develop a comprehensive Bayesian network model to predict noise annoyance, by identifying main contributing factors.MethodsWe conducted a deep literature review to list variables affecting the probability of noise annoyance Expert knowledge was elicited to finalize variables and find their causal relationship. We collected questionnaires from 542 workers in a steel factory during 2025-2023 to gather data on noise annoyance, insomnia severity, general health, workload, personality type, job stress, and loudness perception as the key variables. The Bayesian network was applied to model the probability of noise annoyance among workers and determine its key influencing factors. The model's accuracy and sensitivity were ultimately evaluated based on different approaches.ResultsQuestionnaires results indicated that over 50% of the workers found the noise annoying, and nearly half perceived the loudness as extremely high. BN results showed that stress, sensitivity to noise, and personality were identified as the most influential factors affecting noise annoyance, respectively. The model exhibited a low error rate of 11.81% and a high Area Under the Curve of 0.88, demonstrating a high level of accuracy in predicting noise annoyance levels experienced by workers.ConclusionThe study highlights the developed Bayesian model as a valuable tool for predicting risk factors and emphasizes pre-employment health assessments incorporating stress, noise sensitivity and personality to prevent, identify and manage noise annoyance in industrial settings.