Weighing and modelling factors influencing serum cortisol and melatonin concentration among workers that are exposed to various sound pressure levels using neural network algorithm: An empirical study

• Published in: Heliyon 2020-09, Vol.6 (9), p.e05044-e05044, Article e05044
• Main Authors: Zare, Sajad, Hemmatjo, Rasoul, ElahiShirvan, Hossein, Malekabad, Ashkan Jafari, Kazemi, Reza, Nadri, Farshad
• Format: Article
• Language: English
• Subjects: Acoustical engineering; algorithms; blood serum; computer software; Cortisol; empirical research; Industrial engineering; Melatonin; Modelling; Neural network; Noise; Public health; radioimmunoassays; Safety engineering; Software engineering; Theoretical computer science; working conditions
• ISSN: 2405-8440; 2405-8440
• DOI: 10.1016/j.heliyon.2020.e05044

Noise is one of the most common harmful agents in the workplace. Exposure to excessive noise can lead to complications such as cardiovascular disorders, disturbance of body hormones’ rhythm and hearing loss. This study aimed at weighing and modelling factors influencing serum cortisol and melatonin concentrations of workers that are exposed to various sound pressure levels using neural network algorithm. A case-control design was adopted in the current research. The required data were collected from 75 industrial and mining firm staff members. They were assigned to three groups with equal sample sizes (25 workers). In developing the conceptual model in regard to variables that may affect workers’ serum cortisol and melatonin concentration, SPL, age, weight, and height were included. The influence of SPL on serum cortisol concentration as assessed in the three shifts. Moreover, radioimmunoassay (RIA) was utilized to assess serum cortisol and melatonin concentrations. Neural network algorithm was subsequently exploited to weigh and model predictor factors. IBM SPSS Modeler 18.0 was the software program used for data analysis. The average cortisol concentration values for administrative, condensing, and pelletizing units respectively were 10.24 ± 2.35, 12.15 ± 3.46, and 14.91 ± 4.16μgdl. On the other hand, the average melatonin concentration values for administrative, condensing, and pelletizing units respectively were 37 ± 12.52, 34 ± 13.15, and 27 ± 9.54μgdl. According to the results of the developed model for cortisol, SPL3 (32%) and age (5%) respectively had the highest and lowest impact. On the other hand, considering the model developed for melatonin, height (27%) and SPL1 (10%) were the most and least influential factors in that order. The accuracy rates of the model were also found to be 95% for cortisol and 97% for melatonin. Comparing cortisol concentrations during various shifts revealed a significant reduction (from the beginning to the end of the shift) in all the three groups. Further, the rise of SPL would result in higher secretion of cortisol. Moreover, in all the three groups, the average serum melatonin concentration went up from the beginning to the middle of the shift and then declined to the end of the shift. Considering the accuracy rates of the models developed to predict hormones, neural network algorithm is a suitable and powerful tool for weighing and modelling factors influencing serum cortisol and melatonin concentrations. Industr